You are on page 1of 5

The Power Fit

Sometimes, depending on the nature of the data to be approximated, the


function to be required is in the form 𝑦 = 𝑓(𝑥) = 𝐴𝑥 , for certain known value
of M. Such type of curve fit is called Power Fit. It may be noted that the third
/
law of planetary motion, expressed by 𝑇 = 𝐶𝑥 , 𝐶 = 0.199769 is an example
of Power Fit.

Following is the theorem providing the sufficient condition to minimize the


root-mean-squares error, in this regard.

Theorem:

Consider the data points {(𝑥 , 𝑦 )} for the distinct {𝑥 } . The data points
are approximated by the least-squares power curve 𝑦 = 𝑓(𝑥) = 𝐴𝑥 , where M
is constant. Then, A can be obtained as:

𝐴= 𝑥 𝑦 𝑥

To minimize the root-mean-squares error

/
1
𝐸 (𝑓) = |𝑓(𝑥 ) − 𝑦 | ,
𝑁

it is sufficient to minimize

𝐸(𝐴) = 𝑁[𝐸 (𝑓)] = (𝑓(𝑥 ) − 𝑦 ) = (𝐴𝑥 −𝑦 )

To minimize 𝐸(𝐴), putting 𝐸′(𝐴) = 0, i.e.

𝐸( )
= 2(𝐴𝑥 − 𝑦 )𝑥 =2 (𝐴𝑥 −𝑦 𝑥 )=0

i.e.
𝐴 𝑥 = 𝑥 𝑦

or

(∑ 𝑥 𝑦 )
𝐴=
(∑ 𝑥 )

Exercises:

Q1. Find the least-squares line 𝑦 = 𝑓(𝑥) = 𝐴𝑥 + 𝐵 for the data given in the
following table. Calculate 𝐸 (𝑓) as well.

𝑥 𝑦 𝑓(𝑥 )

-6 -5.3 -6.00

-2 -3.5 -2.84

0 -1.7 -1.26

2 0.2 0.32

6 4.0 3.48

Putting
𝑥 , 𝑓(𝑥 ) = (−6, −6) and 𝑥 , 𝑓(𝑥 ) = (−2, −2.84)
in the least-squares line 𝑦 = 𝑓(𝑥) = 𝐴𝑥 + 𝐵, it can be written as:
−6 = −6𝐴 + 𝐵
−2.84 = −2𝐴 + 𝐵

which leads to the values of A and B as


𝐴 = 0.79 and 𝐵 = −1.26
Hence, the least-squares line is 𝑦 = 𝑓(𝑥) = 0.79𝑥 − 1.26, as shown below
/
To determine 𝐸 (𝑓) = ∑ {|𝑓(𝑥 ) − 𝑦 | } , the following table of
summations is used.

𝑦 𝑓(𝑥 ) 𝑓(𝑥 ) − 𝑦 |𝑓(𝑥 ) − 𝑦 |

-5.3 -6.00 -0.7 0.49

-3.5 -2.84 0.66 0.4356

-1.7 -1.26 0.44 0.1936

0.2 0.32 0.12 0.0144

4.0 3.48 -0.52 0.2704

Sum=1.404
Thus, 𝐸 (𝑓) ≈ 0.5299.

Q2. Find the Power Fit 𝑦 = 𝑓(𝑥) = 𝐴𝑥, where 𝑀 = 1, which is a line passing
through the origin, for the data given in the following table. Calculate 𝐸 (𝑓) as
well.

𝑥 𝑦 𝑓(𝑥 )

3 1.6 1.722

4 2.4 2.296

5 2.9 2.870

6 3.4 3.444

8 4.6 4.592

Putting
𝑥 , 𝑓(𝑥 ) = (3,1.722)
in the Power Fit 𝑦 = 𝑓(𝑥) = 𝐴𝑥, it can be written as:
1.722 = 𝐴3
𝐴 = 0.574

Hence, the Power Fit is


𝑦 = 𝑓(𝑥) = 0.574𝑥

To determine
/
1
𝐸 (𝑓) = {|𝑓(𝑥 ) − 𝑦 | }
5

the following table of summations is used.


𝑦 𝑓(𝑥 ) 𝑓(𝑥 ) − 𝑦 |𝑓(𝑥 ) − 𝑦 |

1.6 1.722 0.122 0.014884

2.4 2.296 -0.104 0.010816

2.9 2.870 -0.03 0.0009

3.4 3.444 0.044 0.001936

4.6 4.592 -0.008 0.000064

Sum=0.0286

Thus, 𝐸 (𝑓) ≈ 0.1689639 .

You might also like