See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320357017

Nonlinear Nonparametric Statistics ("NNS" R-Package)

Method · September 2017

DOI: 10.13140/RG.2.2.28704.48643

CITATION READS

1 129

2 authors, including:

Fred Viole

35 PUBLICATIONS   45 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

All working papers co-authored with Fred Viole constitute my current working projects. View project

All content following this page was uploaded by Fred Viole on 17 May 2019.

The user has requested enhancement of the downloaded file.

 Package ‘NNS’
May 17, 2019
Type Package
Title Nonlinear Nonparametric Statistics
Version 0.4.1
Date 2019-05-15
Author Fred Viole
Maintainer Fred Viole <ovvo.financial.systems@gmail.com>
Description Nonlinear nonparametric statistics using partial moments. Partial moments are the ele-
ments of variance and asymptotically approximate the area of f(x). These robust statistics pro-
vide the basis for nonlinear analysis while retaining linear equivalences. NNS offers: Numeri-
cal integration, Numerical differentiation, Clustering, Correlation, Dependence, Causal analy-
sis, ANOVA, Regression, Classification, Seasonality, Autoregressive modeling, Normaliza-
tion and Stochastic dominance. All routines based on: Viole, F. and Nawrocki, D. (2013), Non-
linear Nonparametric Statistics: Using Partial Moments (ISBN: 1490523995).
License GPL-3
LazyData TRUE
RoxygenNote 6.1.1
Depends R (>= 3.3.0), doParallel
Imports data.table, rgl, stringr
Suggests knitr, rmarkdown
VignetteBuilder knitr

R topics documented:
Co.LPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Co.UPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
D.LPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
D.UPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
dy.dx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
dy.d_ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
LPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
LPM.ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
LPM.VaR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
NNS.ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
NNS.ARMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
NNS.ARMA.optim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1
2 Co.LPM

NNS.boost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
NNS.caus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
NNS.cor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
NNS.cor.hd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
NNS.dep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
NNS.diff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
NNS.FSD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
NNS.FSD.uni . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
NNS.norm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
NNS.part . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
NNS.PDF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
NNS.reg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
NNS.SD.efficient.set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
NNS.seas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
NNS.SSD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
NNS.SSD.uni . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
NNS.stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
NNS.term.matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
NNS.TSD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
NNS.TSD.uni . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
PM.matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
UPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
UPM.ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
UPM.VaR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

Index 44

Co.LPM Co-Lower Partial Moment (Lower Left Quadrant 4)

Description
This function generates a co-lower partial moment for between two equal length variables for any
degree or target.

Usage
Co.LPM(degree.x, degree.y, x, y, target.x = mean(x),
target.y = mean(y))

Arguments
degree.x integer; Degree for variable X. (degree.x = 0) is frequency, (degree.x = 1)
is area.
degree.y integer; Degree for variable Y. (degree.y = 0) is frequency, (degree.y = 1)
is area.
x a numeric vector.
y a numeric vector of equal length to x.
target.x numeric; Typically the mean of Variable X for classical statistics equivalences,
but does not have to be. (Vectorized)
target.y numeric; Typically the mean of Variable Y for classical statistics equivalences,
but does not have to be. (Vectorized)
Co.UPM 3

Value
Co-LPM of two variables

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
Co.LPM(0, 0, x, y, mean(x), mean(y))

Co.UPM Co-Upper Partial Moment (Upper Right Quadrant 1)

Description
This function generates a co-upper partial moment between two equal length variables for any
degree or target.

Usage
Co.UPM(degree.x, degree.y, x, y, target.x = mean(x),
target.y = mean(y))

Arguments
degree.x integer; Degree for variable X. (degree.x = 0) is frequency, (degree.x = 1)
is area.
degree.y integer; Degree for variable Y. (degree.y = 0) is frequency, (degree.y = 1)
is area.
x a numeric vector.
y a numeric vector of equal length to x.
target.x numeric; Typically the mean of Variable X for classical statistics equivalences,
but does not have to be. (Vectorized)
target.y numeric; Typically the mean of Variable Y for classical statistics equivalences,
but does not have to be. (Vectorized)

Value
Co-UPM of two variables

Author(s)
Fred Viole, OVVO Financial Systems
4 D.LPM

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x<-rnorm(100) ; y<-rnorm(100)
Co.UPM(0,0,x,y,mean(x),mean(y))

D.LPM Divergent-Lower Partial Moment (Lower Right Quadrant 3)

Description
This function generates a divergent lower partial moment between two equal length variables for
any degree or target.

Usage
D.LPM(degree.x, degree.y, x, y, target.x = mean(x), target.y = mean(y))

Arguments
degree.x integer; Degree for variable X. (degree.x = 0) is frequency, (degree.x = 1)
is area.
degree.y integer; Degree for variable Y. (degree.y = 0) is frequency, (degree.y = 1)
is area.
x a numeric vector.
y a numeric vector of equal length to x.
target.x numeric; Typically the mean of Variable X for classical statistics equivalences,
but does not have to be. (Vectorized)
target.y numeric; Typically the mean of Variable Y for classical statistics equivalences,
but does not have to be. (Vectorized)

Value
Divergent LPM of two variables

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
D.LPM(0, 0, x, y, mean(x), mean(y))
D.UPM 5

D.UPM Divergent-Upper Partial Moment (Upper Left Quadrant 2)

Description

This function generates a divergent upper partial moment between two equal length variables for
any degree or target.

Usage

D.UPM(degree.x, degree.y, x, y, target.x = mean(x), target.y = mean(y))

Arguments

degree.x integer; Degree for variable X. (degree.x = 0) is frequency, (degree.x = 1)

is area.
degree.y integer; Degree for variable Y. (degree.y = 0) is frequency, (degree.y = 1)
is area.
x a numeric vector.
y a numeric vector of equal length to x.
target.x numeric; Typically the mean of Variable X for classical statistics equivalences,
but does not have to be. (Vectorized)
target.y numeric; Typically the mean of Variable Y for classical statistics equivalences,
but does not have to be. (Vectorized)

Value

Divergent UPM of two variables

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
D.UPM(0, 0, x, y, mean(x), mean(y))
6 dy.dx

dy.dx Partial Derivative dy/dx

Description
Returns the numerical partial derivate of y wrt x for a point of interest.

Usage
dy.dx(x, y, order = NULL, stn = 0.99, eval.point = median(x),
deriv.order = 1, h = 0.05, noise.reduction = "mean",
deriv.method = "FS")

Arguments
x a numeric vector.
y a numeric vector.
order integer; Controls the number of partial moment quadrant means. Defaults to
(order = NULL) which generates a more accurate derivative for well specified
cases.
stn numeric [0, 1]; Signal to noise parameter, sets the threshold of NNS.dep which
reduces "order" when (order = NULL). Defaults to 0.99 to ensure high de-
pendence for higher "order" and endpoint determination.
eval.point numeric; x point to be evaluated. Defaults to (eval.point = median(x)). Set
to (eval.point = "overall") to find an overall partial derivative estimate.
deriv.order numeric options: (1, 2); 1 (default) for first derivative. For second derivative
estimate of f(x), set (deriv.order = 2).
h numeric [0, ...]; Percentage step used for finite step method. Defaults to h = .05
representing a 5 percent step from the value of the independent variable.
noise.reduction
the method of determing regression points options: ("mean", "median", "mode",
"off"); In low signal to noise situations, (noise.reduction = "median") uses
medians instead of means for partitions, while (noise.reduction = "mode")
uses modes instead of means for partitions. (noise.reduction = "off") al-
lows for maximum possible fit in NNS.reg. Default setting is (noise.reduction = "mean").
deriv.method method of derivative estimation, options: ("NNS", "FS"); Determines the partial
derivative from the coefficient of the NNS.reg output when (deriv.method = "NNS")
or generates a partial derivative using the finite step method (deriv.method = "FS")
(Defualt).

Value
Returns the value of the partial derivative estimate for the given order.

Note
If a vector of derivatives is required, ensure (deriv.method = "FS").
dy.d_ 7

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995
Vinod, H. and Viole, F. (2017) "Nonparametric Regression Using Clusters" https://link.springer.
com/article/10.1007/s10614-017-9713-5

Examples
## Not run:
x <- seq(0, 2 * pi, pi / 100) ; y <-sin(x)
dy.dx(x, y, eval.point = 1.75)
## End(Not run)

# Vector of derivatives
## Not run:
dy.dx(x, y, eval.point = c(1.75, 2.5), deriv.method = "FS")
## End(Not run)

dy.d_ Partial Derivative dy/d_[wrt]

Description
Returns the numerical partial derivate of y with respect to [wrt] any regressor for a point of interest.
Finite difference method is used with NNS.reg estimates as f(x + h) and f(x - h) values.

Usage
dy.d_(x, y, wrt, eval.points = "median", order = NULL, stn = 0.99,
h = 0.05, n.best = NULL, mixed = FALSE, plot = FALSE,
noise.reduction = "mean")

Arguments
x a numeric matrix or data frame.
y a numeric vector with compatible dimsensions to x.
wrt integer; Selects the regressor to differentiate with respect to.
eval.points numeric or options: ("mean", median", "last"); Regressor points to be evaluated.
(eval.points = "median") (default) to find partial derivatives at the median
of every variable. Set to (eval.points = "last") to find partial derivatives
at the last value of every variable. Set to (eval.points="mean") to find partial
derivatives at the mean value of every variable. Set to (eval.points = "all")
to find partial derivatives at every observation.
order integer; NNS.reg "order", defaults to NULL.
stn numeric [0, 1]; Signal to noise parameter, sets the threshold of NNS.dep which
reduces "order" when (order = NULL). Defaults to 0.99 to ensure high de-
pendence for higher "order" and endpoint determination.
8 dy.d_

h numeric [0, ...]; Percentage step used for finite step method. Defaults to h = .05
representing a 5 percent step from the value of the regressor.
n.best integer; Sets the number of closest regression points to use in estimating finite
difference points in NNS.reg. NULL (default) Uses ceiling(sqrt(ncol(x))).
mixed logical; FALSE (default) If mixed derivative is to be evaluated, set (mixed = TRUE).
Only for single valued eval.points.
plot logical; FALSE (default) Set to (plot = TRUE) to view plot.
noise.reduction
the method of determing regression points options: ("mean", "median", "mode",
"off"); In low signal to noise situations, (noise.reduction = "median") uses
medians instead of means for partitions, while (noise.reduction = "mode")
uses modes instead of means for partitions. (noise.reduction = "off") al-
lows for maximum possible fit in NNS.reg. Default setting is (noise.reduction = "mean").

Value
Returns:
• dy.d_(...)$"First Derivative" the 1st derivative
• dy.d_(...)$"Second Derivative" the 2nd derivative
• dy.d_(...)$"Mixed Derivative" the mixed derivative (for two independent variables only).
Retuns a vector of partial derivatives when (eval.points = "all").

Note
For known function testing and analysis, regressors should be transformed via expand.grid to fill
the dimensions with (order = "max"). Example provided below.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
## Not run:
set.seed(123) ; x_1 <- runif(100) ; x_2 <- runif(100) ; y <- x_1 ^ 2 * x_2 ^ 2
B = cbind(x_1, x_2)
## To find derivatives of y wrt 1st regressor
dy.d_(B, y, wrt = 1, eval.points = c(.5, .5))
## End(Not run)

## Known function analysis: [y = a ^ 2 * b ^ 2]

## Not run:
x_1 <- seq(0, 1, .1) ; x_2 <- seq(0, 1, .1)
B = expand.grid(x_1, x_2) ; y <- B[ , 1] ^ 2 * B[ , 2] ^ 2
dy.d_(B, y, wrt = 1, eval.points = c(.5, .5), order = "max")
## End(Not run)
LPM 9

LPM Lower Partial Moment

Description
This function generates a univariate lower partial moment for any degree or target.

Usage
LPM(degree, target, variable)

Arguments
degree integer; (degree = 0) is frequency, (degree = 1) is area.
target numeric; Typically set to mean, but does not have to be. (Vectorized)
variable a numeric vector.

Value
LPM of variable

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100)
LPM(0, mean(x), x)

LPM.ratio Lower Partial Moment RATIO

Description
This function generates a standardized univariate lower partial moment for any degree or target.

Usage
LPM.ratio(degree, target, variable)
10 LPM.VaR

Arguments
degree integer; (degree = 0) is frequency, (degree = 1) is area.
target numeric; Typically set to mean, but does not have to be. (Vectorized)
variable a numeric vector.

Value
Standardized LPM of variable

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100)
LPM.ratio(0, mean(x), x)

## Continuous CDF
LPM.ratio(1, sort(x), x)

## Joint CDF
## Not run:
x <- rnorm(5000) ; y <- rnorm(5000)
plot3d(x, y, Co.LPM(0, 0, sort(x), sort(y), x, y), col = "blue", xlab = "X", ylab = "Y",
zlab = "Probability", box = FALSE)

## End(Not run)

LPM.VaR LPM VaR

Description
Generates a VaR based on the Lower Partial Moment ratio.

Usage
LPM.VaR(percentile, degree, x)

Arguments
percentile numeric [0, 1]; The percentile for left-tail VaR.
degree integer; (degree = 0) for discrete distributions, (degree = 1) for continuous
distributions.
x a numeric vector.
NNS.ANOVA 11

Value
Returns a numeric value representing the point at which "percentile" of the area of x is above.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100)

## For 95th percentile VaR (left-tail)

LPM.VaR(0.95, 0, x)

NNS.ANOVA NNS ANOVA

Description
Analysis of variance (ANOVA) based on lower partial moment CDFs for multiple variables. Returns
a degree of certainty the difference in sample means is zero, not a p-value.

Usage
NNS.ANOVA(control, treatment, confidence.interval = 0.95,
tails = "Both", pairwise = FALSE, plot = TRUE, binary = TRUE)

Arguments
control a numeric vector, matrix or data frame.
treatment NULL (default) a numeric vector, matrix or data frame.
confidence.interval
numeric [0, 1]; The confidence interval surrounding the control mean when
(binary = TRUE). Defaults to (confidence.interval = 0.95).
tails options: ("Left", "Right", "Both"). tails = "Both"(Default) Selects the tail of
the distribution to determine effect size.
pairwise logical; FALSE (defualt) Returns pairwise certainty tests when set to pairwise = TRUE.
plot logical; TRUE (default) Returns the boxplot of all variables along with grand
mean identification. When (binary = TRUE), returns the boxplot of both vari-
ables along with grand mean identification and confidence interval thereof.
binary logical; TRUE (default) Selects binary analysis between a control and treatment
variable.
12 NNS.ARMA

Value
For (binary = FALSE) returns the degree certainty the difference in sample means is zero [0, 1].
For (binary = TRUE) returns:
• "Control Mean"
• "Treatment Mean"
• "Grand Mean"
• "Control CDF"
• "Treatment CDF"
• "Certainty" the certainty of the same population statistic
• "Lower Bound Effect" and "Upper Bound Effect" the effect size of the treatment for the
specified confidence interval

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
### Binary analysis and effect size
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.ANOVA(control = x, treatment = y)

### Two variable analysis with no control variable

A <- cbind(x, y)
NNS.ANOVA(A)

### Multiple variable analysis with no control variable

set.seed(123)
x <- rnorm(100) ; y<- rnorm(100) ; z<- rnorm(100)
A <- cbind(x, y, z)
NNS.ANOVA(A)

NNS.ARMA NNS ARMA

Description
Autoregressive model incorporating nonlinear regressions of component series.

Usage
NNS.ARMA(variable, h = 1, training.set = NULL,
seasonal.factor = TRUE, best.periods = 2, negative.values = FALSE,
method = "nonlin", dynamic = FALSE, plot = TRUE,
seasonal.plot = TRUE, intervals = FALSE, ncores = NULL)
NNS.ARMA 13

Arguments
variable a numeric vector.
h integer; 1 (default) Number of periods to forecast.
training.set numeric; NULL (defualt) Sets the number of variable observations
(variable[1 : training.set]) to monitor performance of forecast over in-
sample range.
seasonal.factor
logical or integer(s); TRUE (default) Automatically selects the best seasonal lag
from the seasonality test. To use weighted average of all seasonal lags set to
(seasonal.factor = FALSE). Otherwise, directly input known frequency in-
teger lag to use, i.e. (seasonal.factor = 12) for monthly data. Multiple fre-
quency integers can also be used, i.e. (seasonal.factor = c(12, 24, 36))
best.periods integer; [2] (default) used in conjuction with (seasonal.factor = FALSE),
uses the best.periods number of detected seasonal lags instead of ALL lags
when
(seasonal.factor = FALSE).
negative.values
logical; FALSE (default) If the variable can be negative, set to (negative.values = TRUE).
method options: ("lin", "nonlin", "both"); "nonlin" (default) To select the regression
type of the component series, select (method = "both") where both linear
and nonlinear estimates are generated. To use a nonlineaer regression, set to
(method = "nonlin"); to use a linear regression set to (method = "lin").
dynamic logical; FALSE (default) To update the seasonal factor with each forecast point,
set to (dynamic = TRUE). The default is (dynamic = FALSE) to retain the
original seasonal factor from the inputted variable for all ensuing h.
plot logical; TRUE (default) Returns the plot of all periods exhibiting seasonality and
the variable level reference in upper panel. Lower panel returns original data
and forecast.
seasonal.plot logical; TRUE (default) Adds the seasonality plot above the forecast. Will be set
to FALSE if no seasonality is detected or seasonal.factor is set to an integer
value.
intervals logical; FALSE (default) Plots the surrounding forecasts around the final estimate
when (intervals = TRUE) and (seasonal.factor = FALSE). There are no
other forecasts to plot when a single seasonal.factor is selected.
ncores integer; value specifying the number of cores to be used in the parallelized pro-
cedure. If NULL (default), the number of cores to be used is equal to the number
of cores of the machine - 1.

Value
Returns a vector of forecasts of length (h).

Note
For monthly data series, increased accuracy may be realized from forcing seasonal factors to multi-
ples of 12. For example, if the best periods reported are: {37, 47, 71, 73} use (seasonal.factor = c(36, 48, 72)).
(seasonal.factor = FALSE) can be a very comutationally expensive exercise due to the number
of seasonal periods detected.
14 NNS.ARMA.optim

If error encountered when (seasonal.factor = TRUE):

"NaNs produced Error in seq.default(length(variable)+1, 1, -lag[i]) : wrong sign in 'by' argument"
use the combination of (seasonal.factor = FALSE, best.periods = 1).

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples

## Nonlinear NNS.ARMA using AirPassengers monthly data and 12 period lag

## Not run:
NNS.ARMA(AirPassengers, h = 45, training.set = 100, seasonal.factor = 12, method = "nonlin")
## End(Not run)

## Linear NNS.ARMA using AirPassengers monthly data and 12, 24, and 36 period lags
## Not run:
NNS.ARMA(AirPassengers, h = 45, training.set = 120, seasonal.factor = c(12, 24, 36), method = "lin")
## End(Not run)

## Nonlinear NNS.ARMA using AirPassengers monthly data and 2 best periods lag
## Not run:
NNS.ARMA(AirPassengers, h = 45, training.set = 120, seasonal.factor = FALSE, best.periods = 2)
## End(Not run)

NNS.ARMA.optim NNS ARMA Optimizer

Description

Wrapper function for optimizing any combination of a given seasonal.factor vector in NNS.ARMA.
Minimum sum of squared errors (forecast-actual) is used to determine optimum across all NNS.ARMA
methods.

Usage

NNS.ARMA.optim(variable, training.set, seasonal.factor,

negative.values = FALSE, obj.fn = expression(sum((predicted -
actual)^2)), objective = "min", linear.approximation = TRUE,
depth = 1, print.trace = TRUE)
NNS.ARMA.optim 15

Arguments
variable a numeric vector.
training.set numeric; NULL (defualt) Sets the number of variable observations
seasonal.factor
integers; Multiple frequency integers considered for NNS.ARMA model, i.e.
(seasonal.factor = c(12, 24, 36))
negative.values
logical; FALSE (default) If the variable can be negative, set to (negative.values = TRUE).
obj.fn expression; expression(sum((predicted - actual)^2)) (default) Sum of
squared errors is the default objective function. Any expression() using the
specific terms predicted and actual can be used.
objective options: ("min", "max") "min" (default) Select whether to minimize or maxi-
mize the objective function obj.fn.
linear.approximation
logical; TRUE (default) Uses the best linear output from NNS.reg to generate a
nonlinear and mixture regression for comparison. FALSE is a more exhaustive
search over the objective space.
depth integer; depth = 1 (default) Sets the level from which further combinations are
generated containing only members from prior level’s best seasonal.factors.
print.trace logical; TRUE (defualt) Prints current iteration information. Suggested as backup
in case of error, best parameters to that point still known and copyable!

Value
Returns a list containing a vector of optimal seasonal periods $period, the minimum objective
function value $obj.fn, and the $method identifying which NNS.ARMA method was used.

Note
The number of combinations will grow prohibitively large, they should be kept to a minimum when
(method = "comb").
seasonal.factor containing an element too large will result in an error. Please reduce the maxi-
mum seasonal.factor.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples

## Nonlinear NNS.ARMA period optimization using 2 yearly lags on AirPassengers monthly data
## Not run:
nns.optims <- NNS.ARMA.optim(AirPassengers[1:132], training.set = 120,
seasonal.factor = seq(12, 24, 6))
16 NNS.boost

## End(Not run)

## Then use optimal parameters in NNS.ARMA to predict 12 periods in-sample

## Not run:
NNS.ARMA(AirPassengers, h=12, training.set=132,
seasonal.factor = nns.optims$periods, method = nns.optims$method)

## End(Not run)

NNS.boost NNS Boost

Description
Ensemble method for classification using the predictions of the NNS multivariate regression NNS.reg
collected from uncorrelated feature combinations.

Usage
NNS.boost(IVs.train, DV.train, IVs.test, representative.sample = TRUE,
depth = NULL, n.best = 3, learner.trials = NULL, epochs = NULL,
CV.size = 0.25, threshold = NULL,
obj.fn = expression(mean(round(predicted) == as.numeric(actual))),
objective = "max", extreme = FALSE, feature.importance = TRUE,
status = TRUE, ncores = NULL)

Arguments
IVs.train a matrix or data frame of variables of numeric or factor data types.
DV.train a numeric or factor vector with compatible dimsensions to (IVs.train).
IVs.test a matrix or data frame of variables of numeric or factor data types with compat-
ible dimsensions to (IVs.train).
representative.sample
logical; TRUE (default) Reduces observations of IVs.train to a set of represen-
tative observations per regressor.
depth integer; NULL (default) Specifies the order parameter in the NNS.reg routine,
assigning a number of splits in the regressors. (depth = "max") will be signif-
cantly faster, but increase the variance of results.
n.best integer; 3 (default) Sets the number of nearest regression points to use in weight-
ing for multivariate regression at sqrt(# of regressors). Analogous to k in
a k Nearest Neighbors algorithm.
learner.trials integer; NULL (default) Sets the number of trials to obtain an accuracy threshold
level. Number of observations in the training set is the default setting.
epochs integer; 2*length(DV.train) (default) Total number of feature combinations
to run.
CV.size numeric [0, 1]; (CV.size = .25) (default) Sets the cross-validation size. De-
faults to 0.25 for a 25 percent random sampling of the training set.
threshold numeric [0, 1]; NULL (default) Sets the obj.fn accuracy threshold to keep fea-
ture combinations.
NNS.caus 17

obj.fn expression; expression(mean(round(predicted)==as.numeric(actual)))

(default) Mean accuracy is the default objective function. Any expression()
using the specific terms predicted and actual can be used.
objective options: ("min", "max") "max" (default) Select whether to minimize or maxi-
mize the objective function obj.fn.
extreme logical; FALSE (default) Uses the maximum (minimum) threshold obtained
from the learner.trials, rather than the upper (lower) quintile level for max-
imization (minimization) objective.
feature.importance
logical; TRUE (default) Plots the frequency of features used in the final estimate.
status logical; TRUE (default) Prints status update message in console.
ncores integer; value specifying the number of cores to be used in the parallelized pro-
cedure. If NULL (default), the number of cores to be used is equal to the number
of cores of the machine - 1.

Value
Returns a vector of fitted values for the dependent variable test set.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. (2016) "Classification Using NNS Clustering Analysis" https://ssrn.com/abstract=
2864711

Examples
## Using 'iris' dataset where test set [IVs.test] is 'iris' rows 141:150.
## Not run:
a = NNS.boost(iris[1:140, 1:4], iris[1:140, 5],
IVs.test = iris[141:150, 1:4],
epochs = 100, learner.trials = 100)

## Test accuracy
mean(round(a)==as.numeric(iris[141:150,5]))

## End(Not run)

NNS.caus NNS Causation

Description
Returns the causality from observational data between two variables.

Usage
NNS.caus(x, y, factor.2.dummy = TRUE, tau, time.series = FALSE,
plot = FALSE)
18 NNS.caus

Arguments

x a numeric vector, matrix or data frame.

y NULL (default) or a numeric vector with compatible dimsensions to x.
factor.2.dummy logical; TRUE (default) Automatically augments variable matrix with numerical
dummy variables based on the levels of factors.
tau options: ("cs", "ts", integer); Number of lagged observations to consider (for
time series data). Otherwise, set (tau = "cs") for cross-sectional data. (tau = "ts")
automatically selects the lag of the time series data, while (tau = [integer])
specifies a time series lag.
time.series logical; FALSE (default) If analyzing time series data with tau = [integer],
select (time.series = TRUE). Not required when (tau = "ts").
plot logical; FALSE (default) Plots the raw variables, tau normalized, and cross-
normalized variables.

Value

Returns the directional causation (x —> y) or (y —> x) and net quantity of association. For causal
matrix, directional causation is returned as ([column variable] —> [row variable]). Negative num-
bers represent causal direction attributed to [row variable].

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples

## x clearly causes y...

## Not run:
set.seed(123)
x <- rnorm(100) ; y <- x ^ 2
NNS.caus(x, y, tau = "cs")

x <- 1:100 ; y <- x^2

NNS.caus(x, y, tau = "ts", time.series = TRUE)
## End(Not run)

## Causal matrix
## Not run:
NNS.caus(data.matrix(iris), tau = 0)

## End(Not run)
NNS.cor 19

NNS.cor NNS Correlation

Description
Returns the nonlinear correlation between two variables based on higher order partial moment ma-
trices measured by frequency or area.

Usage
NNS.cor(x, y = NULL, order = NULL, degree = NULL)

Arguments
x a numeric vector, matrix or data frame.
y NULL (default) or a numeric vector with compatible dimsensions to x.
order integer; Controls the level of quadrant partitioning. Defualts to (order = NULL).
Errors can generally be rectified by setting (order = 1).
degree integer; (degree = 0) is frequency based correlations, while (degree = 1)
is for area based correlations. Defaults to (degree = 0) for smaller number of
observations.

Value
Returns nonlinear correlation coefficient between two variables, or nonlinear correlation matrix for
matrix input.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
## Pairwise Correlation
x <- rnorm(100) ; y <- rnorm(100)
NNS.cor(x, y)

## Correlation Matrix
x <- rnorm(100) ; y <- rnorm(100) ; z<-rnorm(100)
B <- cbind(x, y, z)
NNS.cor(B)
20 NNS.cor.hd

NNS.cor.hd NNS Co-Partial Moments Higher Dimension Correlation

Description

Determines higher dimension correlation coefficients based on degree 0 co-partial moments.

Usage

NNS.cor.hd(x, plot = FALSE, independence.overlay = FALSE)

Arguments

x a numeric matrix or data frame.

plot logical; FALSE (default) Generates a 3d scatter plot with regression points using
plot3d.
independence.overlay
logical; FALSE (default) Creates and overlays independent Co.LPM and Co.UPM
regions to visually reference the difference in dependence from the data.frame
of variables being analyzed. Under independence, the light green and red shaded
areas would be occupied by green and red data points respectively.

Value

Returns multivariate nonlinear correlation coefficient

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. (2016) "Beyond Correlation: Using the Elements of Variance for Conditional Means and
Probabilities" http://ssrn.com/abstract=2745308.

Examples

set.seed(123)
x <- rnorm(1000) ; y <- rnorm(1000) ; z <- rnorm(1000)
A <- data.frame(x, y, z)
NNS.cor.hd(A, plot = TRUE, independence.overlay = TRUE)
NNS.dep 21

NNS.dep NNS Dependence

Description
Returns the dependence and nonlinear correlation between two variables based on higher order
partial moment matrices measured by frequency or area.

Usage
NNS.dep(x, y = NULL, order = NULL, degree = NULL,
print.map = FALSE, ncores = NULL)

Arguments
x a numeric vector, matrix or data frame.
y NULL (default) or a numeric vector with compatible dimsensions to x.
order integer; Controls the level of quadrant partitioning. Defaults to (order = NULL).
Errors can generally be rectified by setting (order = 1). Will not partition
further if less than 4 observations exist in a quadrant.
degree integer; Defaults to NULL to allow number of observations to be "degree"
determinant.
print.map logical; FALSE (default) Plots quadrant means.
ncores integer; value specifying the number of cores to be used in the parallelized pro-
cedure. If NULL (default), the number of cores to be used is equal to the number
of cores of the machine - 1.

Value
Returns the bi-variate "Correlation" and "Dependence" or correlation / dependence matrix for
matrix input.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.dep(x, y)

## Correlation / Dependence Matrix

x <- rnorm(100) ; y <- rnorm(100) ; z <- rnorm(100)
B <- cbind(x, y, z)
NNS.dep(B)
22 NNS.diff

NNS.diff NNS Numerical Differentiation

Description

Determines numerical derivative of a given function using projected secant lines on the y-axis.
These projected points infer finite steps h, in the finite step method.

Usage

NNS.diff(f, point, h = 0.1, tol = 1e-10, print.trace = FALSE)

Arguments

f an expression or call or a formula with no lhs.

point numeric; Point to be evaluated for derivative of a given function f.
h numeric [0, ...]; Initial step for secant projection. Defaults to (h = 0.1).
tol numeric; Sets the tolerance for the stopping condition of the inferred h. Defualts
to (tol = 1e-10).
print.trace logical; FALSE (default) Displays each iteration, lower y-intercept, upper y-
intercept and inferred h.

Value

Returns a matrix of values, intercepts, derivatives, inferred step sizes for multiple methods of esti-
mation.

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples

f <- function(x) sin(x) / x

NNS.diff(f, 4.1)
NNS.FSD 23

NNS.FSD NNS FSD Test

Description

Bi-directional test of first degree stochastic dominance using lower partial moments.

Usage

NNS.FSD(x, y, type = "discrete")

Arguments

x a numeric vector.
y a numeric vector.
type options: ("discrete", "continuous"); "discrete" (default) selects the type of
CDF.

Value

Returns one of the following FSD results: "X FSD Y", "Y FSD X", or "NO FSD EXISTS".

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2016) "LPM Density Functions for the Computation of the SD Ef-
ficient Set." Journal of Mathematical Finance, 6, 105-126. http://www.scirp.org/Journal/
PaperInformation.aspx?PaperID=63817.
Viole, F. (2017) "A Note on Stochastic Dominance." https://ssrn.com/abstract=3002675.

Examples

set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.FSD(x, y)
24 NNS.FSD.uni

NNS.FSD.uni NNS FSD Test uni-directional

Description

Uni-directional test of first degree stochastic dominance using lower partial moments used in SD
Efficient Set routine.

Usage

NNS.FSD.uni(x, y, type = "discrete")

Arguments

x a numeric vector.
y a numeric vector.
type options: ("discrete", "continuous"); "discrete" (default) selects the type of
CDF.

Value

Returns (1) if "X FSD Y", else (0).

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2016) "LPM Density Functions for the Computation of the SD Ef-
ficient Set." Journal of Mathematical Finance, 6, 105-126. http://www.scirp.org/Journal/
PaperInformation.aspx?PaperID=63817.
Viole, F. (2017) "A Note on Stochastic Dominance." https://ssrn.com/abstract=3002675.

Examples

set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.FSD.uni(x, y)
NNS.norm 25

NNS.norm NNS Normalization

Description
Normalizes a matrix of variables based on nonlinear scaling normalization method.

Usage
NNS.norm(A, chart.type = NULL, linear = FALSE)

Arguments
A a numeric matrix or data frame.
chart.type options: ("l", "b"); NULL (default). Set (chart.type = "l") for line, (chart.type = "b")
for boxplot.
linear logical; FALSE (default) Performs a linear scaling normalization, resulting in
equal means for all variables.

Value
Returns a data.frame of normalized values.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100) ; y<-rnorm(100)
A <- cbind(x, y)
NNS.norm(A)

NNS.part NNS Partition Map

Description
Creates partitions based on partial moment quadrant means, iteratively assigning identifications to
observations based on those quadrants (unsupervised partitional and hierarchial clustering method).
Basis for correlation NNS.cor, dependence NNS.dep, regression NNS.reg routines.
26 NNS.part

Usage
NNS.part(x, y, Voronoi = FALSE, type = NULL, order = NULL,
max.obs.req = 8, min.obs.stop = FALSE, noise.reduction = "mean")

Arguments
x a numeric vector.
y a numeric vector with compatible dimsensions to x.
Voronoi logical; FALSE (default) Displays a Voronoi type diagram using partial moment
quadrants.
type NULL (default) Controls the partitioning basis. Set to (type = "XONLY") for
X-axis based partitioning. Defaults to NULL for both X and Y-axis partitioning.
order integer; Number of partial moment quadrants to be generated. (order = "max")
will institute a perfect fit.
max.obs.req integer; (8 default) Required observations per cluster where quadrants will not
be further partitioned if observations are not greater than the entered value.
Reduces minimum number of necessary observations in a quadrant to 1 when
(max.obs.req = 1).
min.obs.stop logical; FALSE (default) Stopping condition where quadrants will not be further
partitioned if a single cluster contains less than the entered value of max.obs.req.
noise.reduction
the method of determing regression points options: ("mean", "median", "mode",
"off"); (noise.reduction = "median") uses medians instead of means for
partitions, while (noise.reduction = "mode") uses modes instead of means
for partitions. Defaults to (noise.reduction = "mean"), while (noise.reduction = "off")
will partition quadrant to a single observation for a given (order = ...).

Value
Returns:

• "dt" a data.table of x and y observations with their partition assignment "quadrant" in the
3rd column and their prior partition assignment "prior.quadrant" in the 4th column.
• "regression.points" the data.table of regression points for that given (order = ...).
• "order" the order of the final partition given "min.obs.stop" stopping condition.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.part(x, y)
NNS.PDF 27

## Data.table of observations and partitions

NNS.part(x, y, order = 1)$dt

## Regression points
NNS.part(x, y, order = 1)$regression.points

## Voronoi style plot

NNS.part(x, y, Voronoi = TRUE)

## Examine final counts by quadrant

DT = NNS.part(x, y)$dt
DT[ , counts := .N, by = quadrant]
DT

NNS.PDF NNS PDF

Description

This function generates an empirical PDF using continuous CDFs from LPM.ratio.

Usage

NNS.PDF(variable, degree = 1, target = NULL, bins = NULL,

plot = TRUE)

Arguments

variable a numeric vector.

degree integer; (degree = 0) is frequency, (degree = 1) (default) is area.
target a numeric range of values [a,b] where a < b. NULL (default) uses the variable
observations.
bins numeric; NULL (default) Selects number of observations as default bins.
plot logical; plots PDF.

Value

Returns a data.table containing the intervals used and resulting PDF of the variable.

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995
28 NNS.reg

Examples
set.seed(123)
x <- rnorm(100)
NNS.PDF(x)

## Custom target range

NNS.PDF(x, target = c(-5, 5))

NNS.reg NNS Regression

Description
Generates a nonlinear regression based on partial moment quadrant means.

Usage
NNS.reg(x, y, factor.2.dummy = TRUE, order = NULL, stn = 0.95,
dim.red.method = NULL, tau = NULL, type = NULL, point.est = NULL,
location = "top", return.values = TRUE, plot = TRUE,
plot.regions = FALSE, residual.plot = TRUE, std.errors = FALSE,
confidence.interval = NULL, threshold = 0, n.best = NULL,
noise.reduction = "mean", norm = NULL, dist = "L2",
ncores = NULL, multivariate.call = FALSE)

Arguments
x a vector, matrix or data frame of variables of numeric or factor data types.
y a numeric or factor vector with compatible dimsensions to x.
factor.2.dummy logical; TRUE (default) Automatically augments variable matrix with numerical
dummy variables based on the levels of factors.
order integer; Controls the number of partial moment quadrant means. Users are en-
couraged to try different (order = ...) integer settings with (noise.reduction = "off").
(order = "max") will force a limit condition perfect fit.
stn numeric [0, 1]; Signal to noise parameter, sets the threshold of (NNS.dep) which
reduces ("order") when (order = NULL). Defaults to 0.95 to ensure high
dependence for higher ("order") and endpoint determination.
dim.red.method options: ("cor", "NNS.dep", "NNS.caus", "all", NULL) method for determin-
ing synthetic X* coefficients. Selection of a method automatically engages the
dimension reduction regression. The default is NULL for full multivariate re-
gression. (dim.red.method = "NNS.dep") uses NNS.dep for nonlinear de-
pendence weights, while (dim.red.method = "NNS.caus") uses NNS.caus
for causal weights. (dim.red.method = "cor") uses standard linear correla-
tion for weights. (dim.red.method = "all") averages all methods for further
feature engineering.
tau options("ts", NULL); NULL(default) If (dim.red.method = "NNS.caus") or
(dim.red.method = "all") and the regression is using time-series data, set
(tau = "ts") for more accurate causal analysis.
type NULL (default). To perform a classification, set to (type = "CLASS").
NNS.reg 29

point.est a numeric or factor vector with compatible dimsensions to x. Returns the fitted
value y.hat for any value of x.
location Sets the legend location within the plot, per the x and y co-ordinates used in base
graphics legend.
return.values logical; TRUE (default), set to FALSE in order to only display a regression plot
and call values as needed.
plot logical; TRUE (default) To plot regression.
plot.regions logical; FALSE (default). Generates 3d regions associated with each regression
point for multivariate regressions. Note, adds significant time to routine.
residual.plot logical; TRUE (default) To plot y.hat and Y.
std.errors logical; FALSE (default) To provide standard errors of each linear segment in the
"Fitted.xy" output.
confidence.interval
numeric [0, 1]; NULL (default) Plots the associated confidence interval with the
estimate and reports the standard error for each individual segment.
threshold numeric [0, 1]; (threshold = 0) (default) Sets the threshold for dimension
reduction of independent variables when (dim.red.method) is not NULL.
n.best integer; NULL (default) Sets the number of nearest regression points to use in
weighting for multivariate regression at sqrt(# of regressors). (n.best = "all")
will select and weight all generated regression points. Analogous to k in a
k Nearest Neighbors algorithm. Different values of n.best are tested us-
ing cross-validation in NNS.stack.
noise.reduction
the method of determing regression points options: ("mean", "median", "mode",
"off"); In low signal:noise situations,(noise.reduction = "mean") uses means
for NNS.dep restricted partitions, (noise.reduction = "median") uses medi-
ans instead of means for NNS.dep restricted partitions, while (noise.reduction = "mode")
uses modes instead of means for NNS.dep restricted partitions. (noise.reduction = "off")
allows for maximum possible fit with a specific order.
norm NULL (default) the method of normalization options: ("NNS", "std"); Normalizes
x between 0 and 1 for multivariate regression when set to (norm = "std"), or
normalizes x according to NNS.norm when set to (norm = "NNS").
dist options:("L1", "L2") the method of distance calculation; Selects the distance
calculation used. dist = "L2" (default) selects the Euclidean distance and
(dist = "L1") seclects the Manhattan distance.
ncores integer; value specifying the number of cores to be used in the parallelized pro-
cedure. If NULL (default), the number of cores to be used is equal to the number
of cores of the machine - 1.
multivariate.call
Internal parameter for multivariate regressions.

Value
UNIVARIATE REGRESSION RETURNS THE FOLLOWING VALUES:
• "R2" provides the goodness of fit;
• "SE" returns the overall standard error of the estimate between y and y.hat;
• "Prediction.Accuracy" returns the correct rounded "Point.est" used in classifications
versus the categorical y;
30 NNS.reg

• "derivative" for the coefficient of the x and its applicable range;

• "Point" returns the x point(s) being evaluated;
• "Point.est" for the predicted value generated;
• "regression.points" provides the points used in the regression equation for the given order
of partitions;
• "Fitted" returns a vector containing only the fitted values, y.hat;
• "Fitted.xy" returns a data.table of x,y, y.hat, and NNS.ID;

MULTIVARIATE REGRESSION RETURNS THE FOLLOWING VALUES:

• "R2" provides the goodness of fit;

• "equation" returns the numerator of the synthetic X* dimension reduction equation as a
data.table consisting of regressor and its coefficient. Denominator is simply the length of all
coefficients > 0, returned in last row of equation data.table.
• "x.star" returns the synthetic X* as a vector;
• "rhs.partitions" returns the partition points for each regressor x;
• "RPM" provides the Regression Point Matrix, the points for each x used in the regression
equation for the given order of partitions;
• "Point.est" returns the predicted value generated;
• "Fitted" returns a vector containing only the fitted values, y.hat;
• "Fitted.xy" returns a data.table of x,y, y.hat, gradient, and NNS.ID.

Note

Please ensure point.est is of compatible dimensions to x, error message will ensue if not compati-
ble. Also, upon visual inspection of the data, if a highly periodic variable is observed set (stn = 0)
or (order = "max") to ensure a proper fit.
Identical regressors can be used as long as they do not share the same name. For instance, NNS.reg(cbind(x, 1 * x), y)
will work as NNS.reg is not affected by multicollinearity.
NNS (>= v.0.3.4) has repurposed parameter (type = "CLASS"). (type = "CLASS") is now
restricted to signifying a classification analysis for NNS.reg while (dim.red.method) enables di-
mension reduction regressions.

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995
Vinod, H. and Viole, F. (2017) "Nonparametric Regression Using Clusters" https://link.springer.
com/article/10.1007/s10614-017-9713-5
NNS.reg 31

Examples

## Not run:
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.reg(x, y)
## End(Not run)

## Manual {order} selection

## Not run:
NNS.reg(x, y, order = 2)
## End(Not run)

## Maximum {order} selection

## Not run:
NNS.reg(x, y, order = "max")
## End(Not run)

## x-only paritioning (Univariate only)

## Not run:
NNS.reg(x, y, type = "XONLY")
## End(Not run)

## For Multiple Regression:

## Not run:
x <- cbind(rnorm(100), rnorm(100), rnorm(100)) ; y <- rnorm(100)
NNS.reg(x, y, point.est = c(.25, .5, .75))
## End(Not run)

## For Multiple Regression based on Synthetic X* (Dimension Reduction):

## Not run:
x <- cbind(rnorm(100), rnorm(100), rnorm(100)) ; y<-rnorm(100)
NNS.reg(x, y, point.est = c(.25, .5, .75), dim.red.method = "cor")
## End(Not run)

## IRIS dataset examples:

# Dimension Reduction:
## Not run:
NNS.reg(iris[,1:4], iris[,5], dim.red.method = "cor", order = 5)
## End(Not run)

# Dimension Reduction using causal weights:

## Not run:
NNS.reg(iris[,1:4], iris[,5], dim.red.method = "NNS.caus", order = 5)
## End(Not run)

# Multiple Regression:
## Not run:
NNS.reg(iris[,1:4], iris[,5], order = 2, noise.reduction = "off")
## End(Not run)

# Classification:
## Not run:
NNS.reg(iris[,1:4], iris[,5], point.est = iris[1:10, 1:4], type = "CLASS")$Point.est
## End(Not run)

## To call fitted values:

32 NNS.SD.efficient.set

## Not run:
x <- rnorm(100) ; y <- rnorm(100)
NNS.reg(x, y)$Fitted
## End(Not run)

## To call partial derivative (univariate regression only):

## Not run:
NNS.reg(x, y)$derivative
## End(Not run)

NNS.SD.efficient.set NNS SD Efficient Set

Description
Determines the set of stochastic dominant variables for various degrees.

Usage
NNS.SD.efficient.set(x, degree, type = "discrete")

Arguments
x a numeric matrix or data frame.
degree numeric options: (1, 2, 3); Degree of stochastic dominance test from (1, 2 or 3).
type options: ("discrete", "continuous"); "discrete" (default) selects the type of
CDF.

Value
Returns set of stochastic dominant variable names.

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2016) "LPM Density Functions for the Computation of the SD Ef-
ficient Set." Journal of Mathematical Finance, 6, 105-126. http://www.scirp.org/Journal/
PaperInformation.aspx?PaperID=63817.
Viole, F. (2017) "A Note on Stochastic Dominance." https://ssrn.com/abstract=3002675.

Examples
set.seed(123)
x <- rnorm(100) ; y<-rnorm(100) ; z<-rnorm(100)
A <- cbind(x, y, z)
NNS.SD.efficient.set(A, 1)
NNS.seas 33

NNS.seas NNS Seasonality Test

Description

Seasonality test based on the coefficient of variance for the variable and lagged component series.
A result of 1 signifies no seasonality present.

Usage

NNS.seas(variable, plot = TRUE)

Arguments

variable a numeric vector.

plot logical; TRUE (default) Returns the plot of all periods exhibiting seasonality and
the variable level reference.

Value

Returns a matrix of all periods exhibiting less coefficient of variance than the variable with "all.periods";
and the single period exhibiting the least coefficient of variance versus the variable with "best.period".
If no seasonality is detected, NNS.seas will return ("No Seasonality Detected").

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples

set.seed(123)
x <- rnorm(100)

## To call strongest period based on coefficient of variance:

NNS.seas(x)$best.period
34 NNS.SSD.uni

NNS.SSD NNS SSD Test

Description
Bi-directional test of second degree stochastic dominance using lower partial moments.

Usage
NNS.SSD(x, y)

Arguments
x a numeric vector.
y a numeric vector.

Value
Returns one of the following SSD results: "X SSD Y", "Y SSD X", or "NO SSD EXISTS".

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2016) "LPM Density Functions for the Computation of the SD Ef-
ficient Set." Journal of Mathematical Finance, 6, 105-126. http://www.scirp.org/Journal/
PaperInformation.aspx?PaperID=63817.

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.SSD(x, y)

NNS.SSD.uni NNS SSD Test uni-directional

Description
Uni-directional test of second degree stochastic dominance using lower partial moments used in SD
Efficient Set routine.

Usage
NNS.SSD.uni(x, y)
NNS.stack 35

Arguments
x a numeric vector.
y a numeric vector.

Value
Returns (1) if "X SSD Y", else (0).

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2016) "LPM Density Functions for the Computation of the SD Ef-
ficient Set." Journal of Mathematical Finance, 6, 105-126. http://www.scirp.org/Journal/
PaperInformation.aspx?PaperID=63817.

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.SSD.uni(x, y)

NNS.stack NNS Stack

Description
Prediction model using the predictions of the NNS base models NNS.reg as features (i.e. meta-
features) for the stacked model.

Usage
NNS.stack(IVs.train, DV.train, IVs.test = NULL,
obj.fn = expression(sum((predicted - actual)^2)), objective = "min",
CV.size = NULL, folds = 5, order = NULL, norm = NULL,
method = c(1, 2), dim.red.method = "cor")

Arguments
IVs.train a vector, matrix or data frame of variables of numeric or factor data types.
DV.train a numeric or factor vector with compatible dimsensions to (IVs.train).
IVs.test a vector, matrix or data frame of variables of numeric or factor data types with
compatible dimsensions to (IVs.train).
obj.fn expression; expression(sum((predicted - actual)^2)) (default) Sum of
squared errors is the default objective function. Any expression() using the
specific terms predicted and actual can be used.
objective options: ("min", "max") "min" (default) Select whether to minimize or maxi-
mize the objective function obj.fn.
36 NNS.stack

CV.size numeric [0, 1]; NULL (default) Sets the cross-validation size if (IVs.test = NULL).
Defaults to 0.25 for a 25 percent random sampling of the training set under
(CV.size = NULL).
folds integer; folds = 5 (default) Select the number of cross-validation folds.
order integer; NULL (default) Sets the order for NNS.reg, where (order = 'max') is
the k-nearest neighbors equivalent.
norm options: ("std", "NNS", NULL); NULL (default) 3 settings offered: NULL, "std",
and "NNS". Selects the norm parameter in NNS.reg.
method numeric options: (1, 2); Select the NNS method to include in stack. (method = 1)
selects NNS.reg; (method = 2) selects NNS.reg dimension reduction regres-
sion. Defaults to method = c(1, 2), including both NNS regression methods
in the stack.
dim.red.method options: ("cor", "NNS.dep", "NNS.caus", "all") method for determining syn-
thetic X* coefficients. (dim.red.method = "cor") (default) uses standard
linear correlation for weights. (dim.red.method = "NNS.dep") uses NNS.dep
for nonlinear dependence weights, while (dim.red.method = "NNS.caus")
uses NNS.caus for causal weights. (dim.red.method = "all") averages all
methods for further feature engineering.

Value
Returns a vector of fitted values for the dependent variable test set for all models.

• "NNS.reg.n.best" returns the optimum "n.best" paramater for the NNS.reg multivariate
regression. "SSE.reg" returns the SSE for the NNS.reg multivariate regression.
• "OBJfn.reg" returns the obj.fn for the NNS.reg regression.
• "NNS.dim.red.threshold" returns the optimum "threshold" from the NNS.reg dimension
reduction regression.
• "OBJfn.dim.red" returns the obj.fn for the NNS.reg dimension reduction regression.
• "reg" returns NNS.reg output.
• "dim.red" returns NNS.reg dimension reduction regression output.
• "stack" returns the output of the stacked model.

Note
If character variables are used, transform them first to factors using as.factor, or data.matrix to
ensure overall dataset is numeric. A multifunction sapply can also be applied to the overall dataset:
data <- sapply(data,function(x){as.factor(x) ; as.numeric(x)}). Then run NNS.stack
with transormed variables.
Missing data should be handled prior as well using na.omit or complete.cases on the full dataset.
If error received:
"Error in is.data.frame(x) : object 'RP' not found"
reduce the CV.size.

Author(s)
Fred Viole, OVVO Financial Systems
NNS.term.matrix 37

References
Viole, F. (2016) "Classification Using NNS Clustering Analysis" https://ssrn.com/abstract=
2864711

Examples
## Using 'iris' dataset where test set [IVs.test] is 'iris' rows 141:150.
## Not run:
NNS.stack(iris[1:140, 1:4], iris[1:140, 5], IVs.test = iris[141:150, 1:4])
## End(Not run)

## Using 'iris' dataset to determine [n.best] and [threshold] with no test set.
## Not run:
NNS.stack(iris[ , 1:4], iris[ , 5])
## End(Not run)

## Selecting NNS.reg and dimension reduction techniques.

## Not run:
NNS.stack(iris[1:140, 1:4], iris[1:140, 5], iris[141:150, 1:4], method = c(1, 2))
## End(Not run)

NNS.term.matrix NNS Term Matrix

Description
Generates a term matrix for text classification use in NNS.reg.

Usage
NNS.term.matrix(x, oos = NULL, names = FALSE)

Arguments
x Text A two column dataset should be used. Concatenate text from original
sources to comply with format. Also note the possiblity of factors in "DV",
so "as.numeric(as.character(...))" is used to avoid issues.
oos Out-of-sample text dataset to be classified.
names Column names for "IV" and "oos". Defaults to FALSE.

Value
Returns the text as independent variables "IV" and the classification as the dependent variable "DV".
Out-of-sample independent variables are returned with "OOS".

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995
38 NNS.TSD

Examples
x <- data.frame(cbind(c("sunny", "rainy"), c(1, -1)))
NNS.term.matrix(x)

### Concatenate Text with space seperator, cbind with "DV"

x <- data.frame(cbind(c("sunny", "rainy"), c("windy", "cloudy"), c(1, -1)))
x <- data.frame(cbind(paste(x[ , 1], x[ , 2], sep = " "), as.numeric(as.character(x[ , 3]))))
NNS.term.matrix(x)

### NYT Example

## Not run:
require(RTextTools)
data(NYTimes)

### Concatenate Columns 3 and 4 containing text, with column 5 as DV

NYT=data.frame(cbind(paste(NYTimes[ , 3], NYTimes[ , 4], sep = " "),
as.numeric(as.character(NYTimes[ , 5]))))
NNS.term.matrix(NYT)
## End(Not run)

NNS.TSD NNS TSD Test

Description
Bi-directional test of third degree stochastic dominance using lower partial moments.

Usage
NNS.TSD(x, y)

Arguments
x a numeric vector.
y a numeric vector.

Value
Returns one of the following TSD results: "X TSD Y", "Y TSD X", or "NO TSD EXISTS".

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2016) "LPM Density Functions for the Computation of the SD Ef-
ficient Set." Journal of Mathematical Finance, 6, 105-126. http://www.scirp.org/Journal/
PaperInformation.aspx?PaperID=63817.
NNS.TSD.uni 39

Examples

set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.TSD(x, y)

NNS.TSD.uni NNS TSD Test uni-directional

Description

Uni-directional test of third degree stochastic dominance using lower partial moments used in SD
Efficient Set routine.

Usage

NNS.TSD.uni(x, y)

Arguments

x a numeric vector.
y a numeric vector.

Value

Returns (1) if "X TSD Y", else (0).

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2016) "LPM Density Functions for the Computation of the SD Ef-
ficient Set." Journal of Mathematical Finance, 6, 105-126. http://www.scirp.org/Journal/
PaperInformation.aspx?PaperID=63817.

Examples

set.seed(123)
x <- rnorm(100) ; y <- rnorm(100)
NNS.TSD.uni(x, y)
40 PM.matrix

PM.matrix Partial Moment Matrix

Description
This function generates a co-partial moment matrix for the specified co-partial moment.

Usage
PM.matrix(LPM.degree, UPM.degree, target, variable, pop.adj = FALSE)

Arguments
LPM.degree integer; Degree for variable below target deviations. (degree = 0) is fre-
quency, (degree = 1) is area.
UPM.degree integer; Degree for variable above target deviations. (degree = 0) is fre-
quency, (degree = 1) is area.
target numeric; Typically the mean of Variable X for classical statistics equivalences,
but does not have to be. (Vectorized) (target = "mean") will set the target as
the mean of every variable.
variable a numeric matrix or data.frame.
pop.adj logical; FALSE (default) Adjusts the sample co-partial moment matrices for pop-
ulation statistics.

Value
Matrix of partial moment quadrant values. Uncalled quadrants will return a matrix of zeros.

Note
For divergent asymmetical "D.LPM" and "D.UPM" matrices, matrix is D.LPM(column,row,...).

Author(s)
Fred Viole, OVVO Financial Systems

References
Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples
set.seed(123)
x <- rnorm(100) ; y <- rnorm(100) ; z <- rnorm(100)
A <- cbind(x,y,z)
PM.matrix(LPM.degree = 1, UPM.degree = 1, target = "mean", variable = A)

## Calling Individual Partial Moment Quadrants

cov.mtx = PM.matrix(LPM.degree = 1, UPM.degree = 1, target = "mean", variable = A)
cov.mtx$cupm
UPM 41

## Full covariance matrix

cov.mtx$matrix

UPM Upper Partial Moment

Description

This function generates a univariate upper partial moment for any degree or target.

Usage

UPM(degree, target, variable)

Arguments

degree integer; (degree = 0) is frequency, (degree = 1) is area.

target numeric; Typically set to mean, but does not have to be. (Vectorized)
variable a numeric vector.

Value

UPM of variable

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples

set.seed(123)
x <- rnorm(100)
UPM(0, mean(x), x)
42 UPM.ratio

UPM.ratio Upper Partial Moment RATIO

Description

This function generates a standardized univariate upper partial moment for any degree or target.

Usage

UPM.ratio(degree, target, variable)

Arguments

degree integer; (degree = 0) is frequency, (degree = 1) is area.

target numeric; Typically set to mean, but does not have to be. (Vectorized)
variable a numeric vector.

Value

Standardized UPM of variable

Author(s)

Fred Viole, OVVO Financial Systems

References

Viole, F. and Nawrocki, D. (2013) "Nonlinear Nonparametric Statistics: Using Partial Moments"
http://amzn.com/1490523995

Examples

set.seed(123)
x <- rnorm(100)
UPM.ratio(0, mean(x), x)

## Joint Upper CDF

## Not run:
x <- rnorm(5000) ; y <- rnorm(5000)
plot3d(x, y, Co.UPM(0, 0, sort(x), sort(y), x, y), col = "blue", xlab = "X", ylab = "Y",
zlab = "Probability", box = FALSE)

## End(Not run)
UPM.VaR 43

UPM.VaR UPM VaR

Description
Generates an upside VaR based on the Upper Partial Moment ratio

Usage
UPM.VaR(percentile, degree, x)

Arguments
percentile numeric [0, 1]; The percentile for right-tail VaR.
degree integer; (degree = 0) for discrete distributions, (degree = 1) for continuous
distributions.
x a numeric vector.

Value
Returns a numeric value representing the point at which "percentile" of the area of x is below.

Examples
set.seed(123)
x <- rnorm(100)

## For 95th percentile VaR (right-tail)

UPM.VaR(0.95, 0, x)
Index

∗Topic ANOVA, PM.matrix, 40

NNS.ANOVA, 11 ∗Topic dependence,
∗Topic Autoregressive NNS.cor.hd, 20
NNS.ARMA, 12 NNS.dep, 21
NNS.ARMA.optim, 14 ∗Topic derivative
∗Topic CDF dy.d_, 7
LPM, 9 dy.dx, 6
LPM.ratio, 9 ∗Topic differentiation,
NNS.PDF, 27 NNS.diff, 22
UPM, 41 ∗Topic document
UPM.ratio, 42 NNS.term.matrix, 37
∗Topic Confidence ∗Topic dominance
LPM.VaR, 10 NNS.FSD, 23
∗Topic Interval NNS.FSD.uni, 24
LPM.VaR, 10 NNS.SD.efficient.set, 32
∗Topic PDF, NNS.SSD, 34
NNS.PDF, 27 NNS.TSD, 38
∗Topic VaR, ∗Topic effect
LPM.VaR, 10 NNS.ANOVA, 11
∗Topic VaR ∗Topic matrix
UPM.VaR, 43 NNS.term.matrix, 37
∗Topic causation ∗Topic mean,
NNS.caus, 17 LPM, 9
∗Topic classifier, LPM.ratio, 9
NNS.boost, 16 UPM, 41
∗Topic classifier UPM.ratio, 42
NNS.reg, 28 ∗Topic model
NNS.stack, 35 NNS.ARMA, 12
∗Topic cluster NNS.ARMA.optim, 14
NNS.part, 25 ∗Topic moments,
∗Topic continuous Co.LPM, 2
LPM.ratio, 9 Co.UPM, 3
NNS.PDF, 27 D.LPM, 4
UPM.ratio, 42 D.UPM, 5
∗Topic correlation LPM, 9
NNS.cor, 19 LPM.ratio, 9
NNS.cor.hd, 20 NNS.PDF, 27
NNS.dep, 21 PM.matrix, 40
∗Topic covariance UPM, 41
Co.LPM, 2 UPM.ratio, 42
Co.UPM, 3 ∗Topic multivaiate
D.LPM, 4 dy.d_, 7
D.UPM, 5 ∗Topic nonlinear

44
 INDEX 45

NNS.cor, 19 D.LPM, 4
NNS.reg, 28 D.UPM, 5
∗Topic normalization data.frame, 25
NNS.norm, 25 data.matrix, 36
∗Topic numerical data.table, 26, 30
NNS.diff, 22 dy.d_, 7
∗Topic partial dy.dx, 6
Co.LPM, 2
Co.UPM, 3 expand.grid, 8
D.LPM, 4
D.UPM, 5 legend, 29
dy.d_, 7 LPM, 9
dy.dx, 6 LPM.ratio, 9, 27
LPM, 9 LPM.VaR, 10
LPM.ratio, 9
na.omit, 36
NNS.PDF, 27
NNS.ANOVA, 11
PM.matrix, 40
NNS.ARMA, 12, 14, 15
UPM, 41
NNS.ARMA.optim, 14
UPM.ratio, 42
NNS.boost, 16
∗Topic partitioning,
NNS.caus, 17, 28, 36
NNS.part, 25
NNS.cor, 19, 25
∗Topic regression,
NNS.cor.hd, 20
NNS.reg, 28
NNS.dep, 6, 7, 21, 25, 28, 29, 36
∗Topic regression
NNS.diff, 22
NNS.boost, 16
NNS.FSD, 23
∗Topic seasonality
NNS.FSD.uni, 24
NNS.seas, 33
NNS.norm, 25, 29
∗Topic secant
NNS.part, 25
NNS.diff, 22
NNS.PDF, 27
∗Topic size
NNS.reg, 6–8, 16, 25, 28, 35–37
NNS.ANOVA, 11
NNS.SD.efficient.set, 32
∗Topic stochastic
NNS.seas, 33
NNS.FSD, 23
NNS.SSD, 34
NNS.FSD.uni, 24
NNS.SSD.uni, 34
NNS.SD.efficient.set, 32
NNS.stack, 29, 35
NNS.SSD, 34
NNS.term.matrix, 37
NNS.TSD, 38
NNS.TSD, 38
∗Topic term
NNS.TSD.uni, 39
NNS.term.matrix, 37
∗Topic upper plot3d, 20
UPM, 41 PM.matrix, 40
UPM.ratio, 42
∗Topic variance, sapply, 36
LPM, 9
LPM.ratio, 9 UPM, 41
UPM, 41 UPM.ratio, 42
UPM.ratio, 42 UPM.VaR, 43

as.factor, 36

Co.LPM, 2, 20
Co.UPM, 3, 20
complete.cases, 36

View publication stats