Practice Problems For Curve Fitting - Solution

CME261
Kaiwen Xia
Engineering Mathematics
November 04, 2010
Practice Problems for Curve Fitting - Solution

A. Linear Regression
1. Use least squares regression to fit a straight line to the data given in Table
1. Along with the slope and intercept, compute the standard error of the
estimate and the coefficient of determinant.
Table 1
x
y
0
5
2
6
4
7
6
6
9
9
11
8
2. An investigator has reported the data tabulated in Table 2 for an

experiment to determine the growth rate of bacteria k (per d) as a
function of oxygen concentration (mg/L). It is known that such data can
be modeled by the following equation:
kmax c 2
cs + c 2
where cs and kmax are parameters. Use a transformation to linearize this
equation. Then use linear regression to estimate cs and kmax and predict
the growth rate at c=2 mg/L.
k=
Table 2
C
K
0.5
1.1
0.8
2.4
1.5
5.3
2.5
7.6
4
8.9
B. General Linear Least Squares and nonlinear regression

A physics process can be described with the equation y = f ( x) =
a0 a1
. The
+
x x2
measured values of ( x, y ) are listed in the following table:

Table 3
x
y
1
3
2
0.9
3
0.6
4
0.4
Use direct nonlinear regression method to determine a0 and a1.
CME261
November 04, 2010
Kaiwen Xia
C. Polynomial Interpolation
Assume that the tabulated data in Table 3 are precise; apply the 3rd order Lagrange
Polynomial to approximate f (2.5) with precision to five decimal spaces.
D. Spline interpolation
Fit the data in Table 4 with a quadratic with natural end conditions.
Table 4
x
y
1
55
1.5
22
2
13
2.5
10
CME261
Kaiwen Xia
November 04, 2010
Practice Problems for Curve Fitting - Solution
A. Linear Regression
1. Use least squares regression to fit a straight line to the data given in Table
1. Along with the slope and intercept, compute the standard error of the
estimate and the coefficient of determinant.
Table 3
0
5
x
y
2
6
4
7
6
6
9
9
11
8
Solution:
The straight line that results in the least sum of squares of the residuals
between the measured y and the calculated y will be:
y = a1 x + a0
n xi yi xi yi
where a1 =
n xi2 ( xi )
and a0 = y a1 x
For the data in table 1 we have:

n =6,
= 32 ,
x y
= 41 ,
= 245 ,
2
i
= 258 , x = 5.33 , y = 6.83 .
Substituting in the above relations:

a1 =
6 245 32 41
= 0.30152 and a0 = 6.83 0.30152 5.33 = 5.2258
6 258 322
Therefore, the equation we are looking for is: y = 0.3015 x + 5.2258 .

The standard error of the estimate, SY / X , is given by the following equation:
Sr
n2
where S r is defined as
SY / X =
S r = ei2 = ( yi a0 a1 xi )
n
i =1
i =1
CME261
Kaiwen Xia
November 04, 2010
In this case, Sr is computed to be S r = 2.8931 .
The coefficient of determinant is r2 is the normalized decrease in the error due

to describing the data in terms of a straight line rather than an average value
and it is given by the following equation:
St S r
St
where St is the total sum of squares around the mean for y:
r2 =
St = ( yi y )
i =1
For the given data, we have:
St = 10.833 and r 2 =
10.833 2.893
= 0.7329
10.833
2. An investigator has reported the data tabulated in Table 2 for an

experiment to determine the growth rate of bacteria k (per d) as a
function of oxygen concentration (mg/L). It is known that such data can
be modeled by the following equation:
kmax c 2
cs + c 2
where cs and kmax are parameters. Use a transformation to linearize this
equation. Then use linear regression to estimate cs and kmax and predict
the growth rate at c=2 mg/L.
k=
Table 4
C
K
0.5
1.1
0.8
2.4
1.5
5.3
2.5
7.6
4
8.9
Solution:
The given equation for the growth rate of bacteria can be linearized in the
following way:
kmax c 2
c
1
1
1
= s 2+
2
cs + c
k kmax c kmax
which is of the form y = a1 x + a0
k=
with y = 1/ k , x = 1
c2
, a0 = 1
kmax
and a1 =
cs
kmax
CME261
Kaiwen Xia
November 04, 2010
As a result if we remake the data in table 2 accordingly, we can use linear

regression to estimate cs and kmax.
1
c2
1
k
1.563
0.444
0.160
0.063
0.909
0.417
0.189
0.132
0.112
As in exercise 1, we need to calculate the following quantities for this data:

n = 5,
= 6.299 ,
x = 2.076 , y = 0.586 .
= 1.758 ,
x y
i
= 4.399 ,
2
i
= 18.668 ,
Then we can calculate 0 and 1:
a1 =
5 4.399 6.299 1.758

= 0.202 ,
5 18.668 6.2992
a0 = 0.586 0.202 2.076 = 0.099
Now that we know 0 and 1, we can calculate cs and kmax:
a0 = 1
a1 =
cs
kmax
kmax
kmax = 1
a0
=1
0.099
= 10.1 d-1
cs = a1 kmax = 0.202 10.1 = 2.04 mg/L
and the expression for k is:
k=
10.1 c 2
2.04 + c 2
c = 2 mg / L
k = 6.689 d 1
B. General Linear Least Squares and nonlinear regression

A physics process can be described with the equation y = f ( x) =
a0 a1
. The
+
x x2
measured values of ( x, y ) are listed in the following table:

Table 3
x
y
1
3
2
0.9
3
0.6
4
0.4
Use direct nonlinear regression method to determine a0 and a1.
CME261
Kaiwen Xia
November 04, 2010
S r = ( yi
a0 a1 2
)
xi xi2
1
1
a0 a1
a0 a1
S r
a = 2( yi x x 2 )( x ) = 0
( yi x x 2 )( x ) = 0
0
i
i
i
i
i
i
S
a0 a1
a0 a1
1
1
r
= 2( yi 2 )( 2 ) = 0
( yi x x 2 )( x 2 ) = 0
a1
xi xi
xi
i
i
i
1
1
1 a0 a1
yi
( yi x x 2 x 3 ) = 0
a0 x 2 +a1 x 3 = x
i
i
i
i
i
i
1
1
yi
1 a0 a1
( yi 2 3 4 ) = 0
a0 3 +a1 4 = 2
xi xi xi
xi
xi
xi
i
xi
yi
1
2
3
4
1
2
3
4
3
0.9
0.6
0.4
1 / xi2
1.0000
0.2500
0.1111
0.0625
1.4236
1 / xi3
1.0000
0.1250
0.0370
0.0156
1.1777
1 / xi4
1.0000
0.0625
0.0123
0.0039
1.0788
yi / xi
3.0000
0.4500
0.2000
0.1000
3.7500
yi / xi2
3.0000
0.2250
0.0667
0.0250
3.3167
The system to solve is then:

1.4236a0 + 1.1777 a1 = 3.7500
a = 0.9206
0
a1 = 2.0714
1.1777 a0 + 1.0778a1 = 3.3167
C. Polynomial Interpolation
Assume that the tabulated data in Table 3 are precise; apply the 3rd order Lagrange
Polynomial to approximate f (2.5) with precision to five decimal spaces.
Solution:
The third order Lagrange polynomial for the four data points given in table 3 will be:
4
x xj
i =1
j =1
j i
xi x j
y = yi Li ( x) , Li ( x) =
where:
for i=1
CME261
L1 ( x) =
( x x2 )( x x3 )( x x4 )
( x 2)( x 3)( x 4)
=
( x1 x2 )( x1 x3 )( x1 x4 )
6
( x x1 )( x x3 )( x x4 )
( x 1)( x 3)( x 4)
=
( x2 x1 )( x2 x3 )( x2 x4 )
2
for i=3
L3 ( x) =
Kaiwen Xia
for i=2
L2 ( x) =
November 04, 2010
( x x1 )( x x2 )( x x4 )
( x 1)( x 2)( x 4)
=
( x3 x1 )( x3 x2 )( x3 x4 )
2
and for i=4
L4 ( x) =
( x x1 )( x x2 )( x x3 )
( x 1)( x 2)( x 3)
=
( x4 x1 )( x4 x2 )( x4 x3 )
6
Changing the notation of the data points to make it easier to substitute, gives:
Substituting gives:
f3 ( x) =
( x 2 )( x 3)( x 4 ) 3 + ( x 1)( x 3)( x 4 ) 0.9 +
2
6
( x 1)( x 2 )( x 4 ) 0.6 + ( x 1)( x 2 )( x 3) 0.4
+
6
2
f 3 ( x) = 0.5 ( x 2 )( x 3)( x 4 ) + 0.45 ( x 1)( x 3)( x 4 )

0.3 ( x 1)( x 2 )( x 4 ) + 0.067 ( x 1)( x 2 )( x 3)
To find f(2.5) we substitute x=2.5 in the above equation:
f 3 ( x) = 0.5 ( 2.5 2 )( 2.5 3)( 2.5 4 ) + 0.45 ( 2.5 1)( 2.5 3)( 2.5 4 )
0.3 ( 2.5 1)( 2.5 2 )( 2.5 4 ) + 0.067 ( 2.5 1)( 2.5 2 )( 2.5 3) = 0.63125
CME261
November 04, 2010
Kaiwen Xia
D. Spline interpolation
Fit the data in Table 4 with a quadratic with natural end conditions.
Table 4
x
y
1
55
1.5
22
2
13
2.5
10
Solution:
For quadratic splines, we want to derive a second order polynomial for each interval
which will fit the data in the interval. The polynomial is of the following form:
f i ( x) = ai + bi ( x xi ) + ci ( x xi ) 2
Natural condition implies c1 = 0, b1 = ( y2 y1 ) /( x2 x1 ) .
Use the formula:
ai +1 = yi +1
a1 = 55 a2 = 22 a3 = 13
b1 = 66; b2 = 66; b3 = 30
bi +1 = bi + 2ci ( xi +1 xi )
c = 0 c = 96 c = 72
yi + 2 yi +1
bi +1
3
2
1
ci +1 = ( x x ) 2 ( x x )
i+2
i +1
i+2
i +1
The splines are plotted in the graph below.

60
50
40
30
20
10
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6

Practice Problems For Curve Fitting - Solution

Uploaded by

Copyright:

Available Formats

Practice Problems For Curve Fitting - Solution

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Practice Problems For Curve Fitting - Solution

Uploaded by

Copyright:

Available Formats

CME261

Practice Problems for Curve Fitting - Solution

2. An investigator has reported the data tabulated in Table 2 for an

B. General Linear Least Squares and nonlinear regression

measured values of ( x, y ) are listed in the following table:

Use direct nonlinear regression method to determine a0 and a1.

Practice Problems for Curve Fitting - Solution

For the data in table 1 we have:

= 258 , x = 5.33 , y = 6.83 .

Substituting in the above relations:

Therefore, the equation we are looking for is: y = 0.3015 x + 5.2258 .

The coefficient of determinant is r2 is the normalized decrease in the error due

For the given data, we have:

2. An investigator has reported the data tabulated in Table 2 for an

As a result if we remake the data in table 2 accordingly, we can use linear

As in exercise 1, we need to calculate the following quantities for this data:

Then we can calculate 0 and 1:

5 4.399 6.299 1.758

a0 = 0.586 0.202 2.076 = 0.099

Now that we know 0 and 1, we can calculate cs and kmax:

cs = a1 kmax = 0.202 10.1 = 2.04 mg/L

and the expression for k is:

B. General Linear Least Squares and nonlinear regression

measured values of ( x, y ) are listed in the following table:

Use direct nonlinear regression method to determine a0 and a1.

The system to solve is then:

and for i=4

( x 2 )( x 3)( x 4 ) 3 + ( x 1)( x 3)( x 4 ) 0.9 +

f 3 ( x) = 0.5 ( x 2 )( x 3)( x 4 ) + 0.45 ( x 1)( x 3)( x 4 )

To find f(2.5) we substitute x=2.5 in the above equation:

The splines are plotted in the graph below.

You might also like