Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
42 views

Time Series Analysis

Trend analysis is a statistical technique used to analyze time series data and determine if there are increasing or decreasing trends over time. There are two main types of trends: linear trends represented by a straight line, and nonlinear trends like exponential or polynomial curves. Techniques like regression analysis can be used to fit different models to the data and determine the model with the smallest forecasting error, indicating the best fit.

Uploaded by

Muhammad Nauman
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
42 views

Time Series Analysis

Trend analysis is a statistical technique used to analyze time series data and determine if there are increasing or decreasing trends over time. There are two main types of trends: linear trends represented by a straight line, and nonlinear trends like exponential or polynomial curves. Techniques like regression analysis can be used to fit different models to the data and determine the model with the smallest forecasting error, indicating the best fit.

Uploaded by

Muhammad Nauman
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 9

Trend Analysis

A trend is a long-term component that represents the growth or


decline in the series over an extended period of time. Trend
analysis is a statistical technique to aid interpretation of data.
When a series of measurements of a process are treated as a time
series, trend analysis can be used to make and justify statements
about tendencies in the data.
In particular, it may be useful to determine if measurements exhibit
an increasing or decreasing trend, which is statistically
distinguished from random behavior. Some examples are:
determining the trend of the daily average temperatures at a given
location, from winter to summer; or the trend in a global
temperature series over the last 100 years etc.
Techniques that should be considered when analyzing trend
include regression methods, semi-averages, moving averages,
Holt’s linear exponential smoothing, growth curves and
exponential models etc.
Types of trend
i. Linear trend: A long-term movement of a time series that can
be represented by a straight line. It is usually estimated by
fitting a least square linear equation to the set of observed data
points.
Y^ =a+bt

n ∑ tY −∑Y ∑t
where a=Ý −b t́ and b=
n ∑t 2 −¿ ¿ ¿ .
ii. Nonlinear trend: In some instances, the straight line is not
enough to explain the trend in observed time series. For
example, U.S. GDP per capita has grown at an exponential rate
of approximately two percent per year for two centuries, so
clearly straight line is not appropriate for U.S. GDP. Similarly,
it is always better to model population growth by logistic
curve. We will study two different types of non-linear trends.

i. Polynomial trend :
Polynomial curves fitting points generated with a sine function.
Let us start with a first degree polynomial equation:

This is a line with slope a. We know that a line will connect any
two points. So, a first degree polynomial equation is an exact fit
through any two points with distinct x coordinates.
If we increase the order of the equation to a second degree
polynomial, we get:
This will exactly fit a simple curve to three points.
If we increase the order of the equation to a third degree
polynomial, we get:

This will exactly fit four points.

ii. Exponential trend: It occurs when the observed time


series increases at an exponential rate. For example, the
population growth is usually modeled by an exponential
type curve.
The graph illustrates how exponential growth surpasses both linear
and cubic growth.

Usually the exponential trend is modeled by


Y^ =a bt

This model can be transformed to linear one by using logarithms


i.e.
log ( Y^ ) =log ( a ) +t log (b)

and then simple linear regression determines log ⁡(a) and log(b).
Taking antilogarithm will determine a and b.

Selection of the model:

One way to select an appropriate model is to measure the forecast


error. Smaller error of the model means the better fit to observed
data set.

Measuring Forecasting Error: The difference between observed


values and forecast value is called residual or forecasting error.
That is
e t =Y t−Y^t .
Some of the formulae for evaluating forecasting errors are as
follows:

Mean Absolute Deviation


n
1
MAD= ∑ |Y t −Y^t|
n i=1

Mean Squared Error


n
1 2
MSE= ∑ ( Y t−Y^t )
n i=1

Mean Absolute Percentage Error


n
1 |Y t−Y^t|
MAPE= ∑
n i=1 Y t
Mean Percentage error
n
1 (Y t −Y^t )
MPE= ∑
n i=1 Yt

You might also like