MEAN VARIATION AND STANDARD DEVIATION

If there are n data y1, y2, y3, ... .., yn, then
∑ni=1 yi
Mean y =
n
variation vi = (yi - y)

∑ni=1 (yi - y)2

rms value (Standard deviation) =
n
INTERPOLATION - GENERAL METHOD

We have two points ( x1, y1) and ( x2, y2), we can have a
unique straight line passing through the two points or First
Degree polynomial. We can solve it by the following way.
y1 = a0 + a1 * x1 - (1)
y2 = a0 + a1 * x2 - (2)
y1 - y 1
where, a0 = y1 - a1 * x1 and a1 =  
x1 - x 1

This can be written in Matrix form

y1 1 x1 a
  =   0 
y2 1 x2 a1
a0 1 x1 -1 y
 = =   * 1 
a1 1 x2 y2
Similarlly for three points ( x1, y1), ( x2, y2) and ( x3, y3) .
y1 = a 0 + a 1 * x 1 + a 2 * x 1 2
y2 = a 0 + a 1 * x 2 + a 2 * x 2 2
y3 = a 0 + a 1 * x 3 + a 2 * x 3 2
y1
y2 =
y3
1 x1 x12 a0 a0 1 x1 x12 -1
y1
1 x2 x22 a1 and a1 = 1 x2 x22 y2
1 x3 x32 a2 a2 1 x3 x32 y3
2 2019_Interpolation.nb

Generalizing this →
if we have (n + 1) points ( x0, y0), ( x1, y1), ( x2, y2), ... ..
( x n , yn ) .
we can find the parameters of the unique n th
degree polynomial by solving the matrix equation.
y0 1 x0 x0 2 . . x 0 n a0
y2 1 x1 x1 2 . . x 1 n a1
y3 . . . a2
. = * .
. 1 xn xn 2 . . x n n .
yn an
a0 1 x0 x0 2 . . x 0 n y0
a1 1 x1 x1 2 . . x 1 n y2
a2 . . . y3
= Inv *
. .
. 1 xn xn 2 . . x n n .
an yn

There are novel techniques developed by Newton, Lagrange &

Bessel et. al. which are based on the difference of the data.
Difference

( x0, y0), ( x1, y1), ( x2, y2), ... ..

( xn, yn) denote a set of values of any function y =
f[x]. Then Δ y0 = y1 - y0, Δ y1 = y2 - y1, ... are called
first diagonal differences. The differences
of these first difference (ex : Δ2 y0 = Δ y2 - Δ y1 )
is called second diagonal difference etc ..
Diagonal Difference Table
x y Δy Δ2
y Δ3 y Δ4 y Δ5 y Δ6 y 0
 2019_Interpolation.nb 3

x0 y0 Δ y0 Δ2 y0 Δ3 y0 Δ4 y0 Δ5 y0
x1 y1 Δ y1 Δ2 y1 Δ3 y1 Δ4 y1
x2 y2 Δ y2 Δ2 y2 Δ3 y2
x3 y3 Δ y3 Δ2 y3
x4 y4 Δ y4
x5 y5
The relation between the differences and the y values
x y Δy Δ2
y Δ3 y Δ4 y
x0 y0 y1 - y 0 y2 - 2 y 1 + y 0 y3 -
3 y 2 + 3 y1 - y 0 y4 - 4 y 3 + 6 y 2 - 4 y 1 + y 0
x0 + h y1 y2 - y 1 y3 - 2 y 2 + y 1 y4 -
3 y 3 + 3 y2 - y 1 y5 - 4 y 4 + 6 y 3 - 4 y 2 + y 1
x0 + 2 h y2 y3 - y 2 y4 -
2 y3 + y2 y5 - 3 y 4 + 3 y 3 - y 2
x0 + 3 h y3 y4 - y 3 y5 - 2 y 4 + y 3
x0 + 4 h y4 y5 - y 4
x0 + 5 h y5
Δn yu = Δn-1 yu+1 - Δn-1 yu
From the above tables it can be seen that coefficients with
sign of any difference can be generated from the expansion of
(a - b)n or from the knowledge of nCr. Examples are given below
(a - b)2 = a2 - 2 a b + b2 Δ2 y0 = y2 -2 y1
+y0 (Start from right side y0, y1, y2 with coefficients )
(a - b)3 = a3 -3 a2 b +3 a b2 -b3 so Δ 2 y2 =
y5 - 3 y4 + 3 y3 - y2 (2 - 5 , 4 terms)
(a - b)4 = a4 - 4 a3 b + 6 a2 b2 - 4 a b3 + b4
(a - b)5 = a5 - 5 a4 b + 10 a3 b2 - 10 a2 b3 + 5 a b4 - b5
Horizontal Difference Table
4 2019_Interpolation.nb

( x0, y0), ( x1, y1), ( x2, y2), ... ..

( xn, yn) denote a set of values of any function y =
f[x]. Then Δ1 y1 = y1 - y0, Δ1 y2 = y2 - y1, ... are called
first horizontal differences. The differences
of these first difference (ex : Δ2 y2 = Δ y2 - Δ y1 )
is called second horizontal difference etc ..
x y Δ1 y
Δ2 y Δ3 y Δ4 y Δ5 y
x0 y0
x1 y1 Δ1 y1
x2 y2 Δ1 y2 Δ2 y2
x3 y3 Δ1 y3 Δ2 y3 Δ3 y3
x4 y4 Δ1 y4 Δ2 y4 Δ3 y4 Δ4 y4
x5 y5 Δ1 y5
Δ2 y5 Δ3 y5 Δ4 y5 Δ5 y
The relation between the differences and the y values
x y Δ1 y Δ2 y
Δ3 y Δ4 y Δ5 y
x0 y0
x1 y1 y1 - y 0
x2 y2 y2 - y 1 y2 - 2 y 1 + y 0
x3 y3 y3 - y 2 y3 - 2 y 2 + y 1 y3 - 3 y 2 + 3 y 1 - y 0
x4 y4 y4 - y 3 y4 - 2 y 3 + y 2 y4 -
3 y 3 + 3 y2 - y 1 y4 - 4 y 3 + 6 y 2 - 4 y 1 + y 0
x5 y5 y5 - y 4 y5 - 2 y 4 + y 3 y5 -
3 y 4 + 3 y3 - y 2 y5 - 4 y 4 + 6 y 3 - 4 y 2 +
y1 y5 - 5 y4 + 10 y3 - 10 y2 + 5 y1 - y0
Δn yu = Δn yu - Δn-1 yu-1
The coefficients with sign of any difference can
be generated from the expansion of (a - b)n. Examples
(a - b)2 = a2 - 2 a b + b2
 2019_Interpolation.nb 5

(a - b)3 = a3 -3 a2 b +3 a b2 -b3 so Δ 2 y3 =
y3 - 3 y2 + 3 y1 - y0 (Start from left 3 - 0, 4 terms )
Difference of a polynomial
Let us now compute the succesive
differences of a polynomial of nth degree
y = a0 + a1 x + a2 x2 + a3 x3 + ... ... ....+an xn
y + Δy = f[x + Δx] =
a0 + a1 (x + Δx) + a2 (x + Δx)2 + a3 (x + Δx)3 + ... ... ....+an
(x + Δx)n
h = Δx
Δy =
y + Δy - y = a1 h + a2 ((x + h)2 - x2) + a2 ((x + h)3 - x3) + ... .. +
an ((x + h)n - xn)
(x + h)n - xn =
hn + nC1 hn-1 x + nC2 hn-2 x2 + nC3 hn-3 x3 + ... .. nC1 h xn-2 +
n h xn-1 + x5 - x5
Now the 1 st difference of a n
th degree polynomial is of degree (n - 1)
2 nd difference of n th
degree polynomial is of degree (n - 2)
n th difference of n th degree
polynomial is of degree (n - n) = a constant
Newton' Forward Interpolation formula

Let y = f[x], denotes a function over a given

intervel (whose exact functional form we do not know )
represented by equidistant points ( x0, y0),
( x1, y1), ( x2, y2), ( x3, y3) ... .. ( xn, yn)
ϕ[x] denote a polynomial of nth degree,
the polynomial may be written in the form
6 2019_Interpolation.nb

ϕ[x] =
a0 + a1 (x - x0) + a2 (x - x0) (x - x1) + a3 (x - x0) (x - x1) (x - x2)
+a4 (x - x0) (x - x1) (x - x2) (x - x3)
+ ... ....+an (x - x0) (x - x1) (x - x2) (x - x3) ....(x - xn-1)
We have to find the coefficients a0, a1,
a2 , ... an so as to make ϕ[x0] = y0, ϕ[x1] = y1, .... ϕ[xn] =
yn (or the polynomial is going through all the given points)
Putting x = x0, ϕ[x0] = y0 = a0 + a1 (x0 - x0) + ... ⧴ a0 = y0
Putting x = x1,
y1 - y 0 Δy0
ϕ[x1] = y1 = a0 + a1 (x1 - x0) + ... ⧴ a1 = =
x1 - x 0 h
Putting x = x2,
ϕ[x2] = y2 = a0 + a1 (x1 - x0) + a2 (x2 - x0) (x 2 - x1) ⧴ a2 =
y2 - 2 y 1 + y 0 Δ2 y0
=
2 h2 2 h2
Putting x = x3,
ϕ[x3] = y3 = a0 + a1 (x1 - x0) + a2 (x2 - x0) (x 2 - x1) +
a3 (x3 - x0) (x3 - x1) (x3 - x2)
Δy0 Δ2 y0
y3 = y0 + h+ 2
2 h2 + a3 3.2.h3
h 2h
a3 3.2.h3 = y3 - y0 - Δy0 - Δ2 y0 =
y3 - y0 - Δy0 - Δ2 y0 = y3 - y0 - (y1 - y0) - (y2 - 2 y1 + y0)
y3 - 3 y 2 + 3 y 1 - y 0 Δ3 y0
a3 = =
3 ! h3 3 ! h3
y4 - 4 y 3 + 6 y 2 - 4 y 1 - y 0 Δ4 y0
a4 = = ... ... ... ... ....
4 ! h3 4 ! h3
Δn y0
an =
n ! hn
Substituting the values of parameters
 2019_Interpolation.nb 7

ϕ[x] =
Δy0 Δ2 y0 Δ3 y0
y0 + (x - x0) + (x - x 0 ) (x - x 1 ) + (x - x0) (x - x1)
h 2 h2 3 ! h3
Δ4 y0
(x - x2) + (x - x0) (x - x1)
4 ! h4
Δn y0
(x - x2) (x - x3) + ... ....+
n ! hn
(x - x0) (x - x1) (x - x2) (x - x3) ....(x - xn-1)
This is the Newtons formula for forward interpolation

(x - x0) Δ2 y0 (x - x0) (x - x1)

ϕ[x] = y0 + Δy0 + +
h 2! h h
Δ3 y0 (x - x0) (x - x1) (x - x2) Δ4 y0 (x - x0)
+
3! h h h 4! h
(x - x1) (x - x2) (x - x3) n
Δ y0
+ ... ....+
h h h n!
(x - x0) (x - x1) (x - x2) (x - x3) (x - xn-1)
....
h h h h h
x - x0
Put u =
h
Changing the other values in terms of u,
x - x1 = x - (x0 + h) = x - x0 - h, x - x2 = x - x0 - 2 h, etc.
x - xn = x - (x0 + n * h) = x - x0 - n h
(x - x0) (x - x1) (x - x0) - h
Now put u = , then = = u-1
h h h
(x - x2) (x - x0) - 2 h
= = u-2
h h
Δ2 y0 Δ3 y0
ϕ[x] = y0 + u Δy0 + u (u - 1) + u (u - 1) (u - 2)
2! 3!
4
Δ y0
+u (u - 1) (u - 2) (u - 3) + ... ....+u
4!
Δn y0
(u - 1) (u - 2) (u - 3) ... (u - n + 1)
n!
Newton' Backward Interpolation formula
8 2019_Interpolation.nb

Let y = f[x] denotes a function over a given

intervelrepresented by equidistant points ( x0, y0),
( x1, y1), ( x2, y2), ( x3, y3) ... .. ( xn, yn)
ϕ[x] denote a polynomial of nth degree,
the polynomial may be written in the form
ϕ[x] = a0 + a1 (x - xn) + a2 (x - xn) (x - xn-1)
+a3 (x - xn) (x - xn-1) (x - xn-2) +
a4 (x - xn) (x - xn-1) (x - xn-2) (x - xn-3) + ... ....+an
(x - xn) (x - xn-1) (x - xn-2) (x - xn-3) ....(x - x1)
Determine the coefficients a0, a1,
a2 , ... an so as to make ϕ[x1] = y1, ϕ[x2] = y2, .... ϕ[xn] = yn
Putting x = xn, ϕ[xn] = a0 ... = y0 ⧴ a0 = yn
Putting x = xn-1,
yn - yn-1 Δ1 yn
ϕ[xn-1] = a0 + a1 (xn-1 - xn) + ... = y1 ⧴ a1 = =
xn - xn-1 h
Putting x = xn-2,
ϕ[xn-2] = a0 + a1 (xn-2 - xn) + a2 (xn-2 - xn) (x n-2 - xn-1) =
yn - 2 yn-1 + yn-2 Δ2 y0
yn-2 ⧴ a2 = =
2 h2 2 h2
Putting x = xn-3,
ϕ[xn-3] = a0 + a1 (xn-3 - xn) + a2 (xn-3 - xn) (x n-3 - xn-1) +
a3 (xn-3 - xn) (xn-3 - xn-1) (xn-3 - xn-2) = yn-3
yn - 3 yn-1 + 3 yn-2 - yn-3 Δ3 y0
⧴ a3 = =
3 ! h3 3 ! h3
y4 - 4 y 3 + 6 y 2 - 4 y 1 - y 0 Δ4 yn
a4 = = ... ... ... ... ....
4 ! h3 4 ! h3
Δn yn
an =
n ! hn
Substituting the values of parameters
 2019_Interpolation.nb 9

ϕ[x] =
Δ1 yn Δ2 y0 Δ3 y0
yn + (x - xn) + (x - x n ) (x - x n-1 ) + (x - xn)
h 2 ! h2 3 ! h3
Δ4 yn
(x - xn-1) (x - xn-2) + (x - xn) (x - xn-1)
4 ! h3
Δn yn
(x - xn-2) (x - xn-3) + ... ....+
n ! hn
(x - xn) (x - xn-1) (x - xn-2) (x - xn-3) ....(x - x1)

ϕ[x] =
(x - xn) Δ2 y0 (x - xn) (x - xn-1) Δ3 y0 (x - xn)
yn + Δ 1 yn + +
h 2! h h 3! h
(x - xn-1) (x - xn-2) Δ4 yn (x - xn) (x - xn-1)
+
h h 4! h h
(x - xn-2) (x - xn-3) Δn yn
+ ... ....+
h h n!
(x - xn) (x - xn-1) (x - xn-2) (x - xn-3) (x - x1)
....
h h h h h
x - xn
Put u =
h
x - xn-1
Change the values in terms of u =
h
x - xn + (xn - xn-1)
= u+1
h
x - xn-2
= u + 2 , etc.
h
ϕ[x] =
Δ2 y0 Δ3 y0
yn + u Δ1 yn + u (u + 1) + u (u + 1) (u + 2) + u (u + 1)
2! 3!
Δ4 yn
(u + 2) (u + 3) + ... ....+u
4!
Δn yn
(u + 1) (u + 2) (u + 3) .. (u + n - 1) .
n!
INTERPOLATION - UNEQUAL INTERVALS
10 2019_Interpolation.nb

Interpolation with Unequal Intervals of the

Arguments, sometimes getting values of a function at
equidistant values of its argument is difficult or almost
impossible. Example if we want to standardize a thermocouple
(finding the thermo e.m.fs at some standard temperature ).
We keep it at the melting / boiling points of different
compounts like water, liquid nitrogen etc ..
Let y = f[x], denotes a function over a given intervel
represented by points at unequally distant in its
argument : ( x0, y0), ( x1, y1), ( x2, y2), ( x3, y3) ... ( xn, yn)
First order divided Difference
y1 - y 0 y2 - y 1
= δ (x1, x0), = δ (x2, x1),
x1 - x 0 x2 - x 1
y3 - y 2
= δ (x3, x2) etc ..
x3 - x 2
Second order divided Difference
δ (x2, x1) - δ (x1, x0)
= δ (x2, x1, x0),
x2 - x 0
δ (x3, x2) - δ (x2, x1)
= δ (x3, x2, x1), etc.
x3 - x 1
Third order difference
δ (x4, x3, x2) - δ (x3, x2, x1)
= δ (x4, x3, x2, x1 )
x4 - x 1
Fourth order difference
δ (x4, x3, x2, x1) - δ (x3, x2, x1, x0)
= δ (x4, x3, x2, x1, x0)
x4 - x 0
(n - 1) th order difference
1
(δ (xn-1, xn-2, xn-3, xn-4 ...., x1 ) -
xn-1 - x0
δ (xn-2, xn-3, xn-4, xn-5 ....x1, x0 )) =
δ (xn-1, xn-2, xn-3, xn-4 ...., x0)
n th order difference
 2019_Interpolation.nb 11

1
(δ (xn, xn-1, xn-2, xn-3, xn-4 ...., x1 ) -
xn-1 - x0
δ (xn-1, xn-3, xn-4, xn-5 ....x1, x0 )) =
δ (xn, xn-1, xn-2, xn-3, xn-4 ...., x0)
Symmetry of Divided Difference
y1 - y 0 y0 - y 1
= , Hence δ (x1, x0) = δ (x0, x1)
x1 - x 0 x0 - x 1
y1 y0
δ (x1, x0) = + , In the Denomenator
x1 - x 0 x0 - x 1
starting with the ith argument of the numerator.
δ (x2, x1) - δ (x1, x0)
δ (x2, x0, x1) = =
x2 - x 0
1 y2 y1 y1 y0
 + - + 
(x2 - x0) x2 - x1 x1 - x2 x1 - x 0 x0 - x 1
δ (x2, x0, x1) =
y2 y1 y0
+ +
(x2 - x0) (x2 - x1) (x1 - x0) (x1 - x2) (x0 - x1) (x0 - x2)
δ (x, x0, x1, x2, x3 ....xn-1, xn) =
y / ((x - x0) (x - x1) (x - x2) ....(x - xn-1) (x - xn)) +
y0 / ((x0 - x) (x0 - x1) (x0 - x2) ....(x0 - xn-1)
(x0 - xn)) + ... ... +
yn / (xn - x) (xn - x0) (xn - x1) (xn - x2) ....(xn - xn-1)
The r elation between the differences and the y values
x y Δy Δ2
y Δ3 y Δ4 y
12 2019_Interpolation.nb

x0 y0 y1 - y 0 y2 - 2 y 0 + y 0 y3 -
3 y 2 + 3 y1 - y 0 y4 - 4 y 3 + 6 y 2 - 4 y 1 + y 0

x0 + h y1 y2 - y 1 y3 - 2 y 1 + y 0 y4 -
3 y 3 + 3 y2 - y 1 y5 - 4 y 4 + 6 y 3 - 4 y 2 + y 1

x0 + 2 h y2 y3 - y 2 y4 - 2 y 2 + y 0 y5 - 3 y 4 + 3 y 3 - y 2

x0 + 3 h y3 y4 - y 3 y5 - 2 y 4 + y 3

x0 + 4 h y4 y5 - y 4

x0 + 5 h y5
x y δy δ2
y δ3 y δ4 y
y1 - y 0 y2 - 2 y 0 + y 0
x0 y0
h 2 h. h
y3 - 3 y 2 + 3 y 1 - y 0 y4 - 4 y 3 + 6 y 2 - 4 y 1 + y 0
3h 2h h 4h3h 2h h

y2 - y 1 y3 - 2 y 1 + y 0
x0 + h y1
h 2 h. h
y4 - 3 y 3 + 3 y 2 - y 1 y5 - 4 y 4 + 6 y 3 - 4 y 2 + y 1
3h 2h h 4h3h 2h h

y3 - y 2
x0 + 2 h y2
h
y4 - 2 y 2 + y 0 y5 - 3 y 4 + 3 y 3 - y 2
2 h. h 3h 2h h

y4 - y 3 y5 - 2 y 4 + y 3
x0 + 3 h y3
h 2 h. h

y5 - y 4
x0 + 4 h y4
h

x0 + 5 h y5
 2019_Interpolation.nb 13

Newton' s General Interpolation Formula

y - y0
From = δ (x, x0),
x - x0
rewriting the equation we will get y = y0 + (x - x0) * δ (x, x0)
δ (x, x0) - δ (x0, x1)
δ (x, x0, x1) = , So,
(x - x1)
δ (x, x0) = δ (x0, x1) + (x - x1) * δ (x, x0, x1)
Hence y = y0 + (x - x0) * δ (x, x0) =
y0 + (x - x0) (δ (x0, x1) + (x - x1) * δ (x, x0, x1))
δ (x, x0) = δ (x0, x1) + (x - x1) * δ (x, x0, x1)
y = y0 + (x - x0) (δ (x0, x1) + (x - x1) * δ (x, x0, x1))
y = y0 + (x - x0) δ (x0, x1) + (x - x0) (x - x1) δ (x, x0, x1)
δ (x, x0, x1) = δ (x0, x1, x2) + (x - x2) * δ (x, x0, x1, x2)
y = y0 + (x - x0) * δ (x0, x1) +
(x - x0) * (x - x1) * (δ (x0, x1, x2) + (x - x2) * δ (x, x0, x1, x2))
y=
y0 + (x - x0) δ (x0, x1) + (x - x0) (x - x1) δ (x0, x1, x2) + (x - x0)
(x - x1) (x - x2) δ
(x0, x1, x2, x3) + ... .... ... +
(x - x0) (x - x1) (x - x2) (x - x3) ... (x - xn-1)
δ (x0, x1, x2, x3, x4, ... .., xn-1, xn) +
(x - x0) (x - x1) (x - x2) (x - x3) ... (x - xn-1)
(x - xn) δ (x, x0, x1, x2, x3, x4, ... .., xn-1, xn)

LAGRANGE' S INTERPOLATION FORMULA

nth simple difference of a polynomial of nth

degree is constant. Simillarly the case with
divided difference of a polynomial. Hence (n + 1)th
divided difference of a polynomial is zero.
14 2019_Interpolation.nb

δ (x, x0, x1, x2, x3 ....xn-1, xn) =

y / ((x - x0) (x - x1) (x - x2) ....(x - xn-1) (x - xn)) +
y0 / ((x0 - x) (x0 - x1) (x0 - x2) ....(x0 - xn-1)
(x0 - xn)) + ... ... +
yn / (xn - x) (xn - x0) (xn - x1) (xn - x2) ....(xn - xn-1)
y / ((x - x0) (x - x1) (x - x2) ....(x - xn-1) (x - xn)) =
y0 / ((x0 - x) (x0 - x1) (x0 - x2) ....(x0 - xn-1) (x0 - xn)) +
y1 / ((x1 - x) (x1 - x0) (x1 - x2) ....(x1 - xn-1)
(x1 - xn)) ... ... +
yn / (xn - x) (xn - x0) (xn - x1) (xn - x2) ....(xn - xn-1)
Solving for y
y = (((x - x1) (x - x2) ....(x - xn-1) (x - xn)) /
((x0 - x1) (x0 - x2) ....(x0 - xn-1) (x0 - xn))) y0 +
(((x - x0) (x - x2) ....(x - xn-1) (x - xn)) /
((x1 - x0) (x1 - x2) ....(x1 - xn-1) (x1 - xn))) y1 +
(((x - x0) (x - x1) ....(x - xn-1) (x - xn)) /
((x2 - x0) (x2 - x2) ....(x2 - xn-1) (x2 - xn)))
y2 ... ... + ((x - x0) (x - x1) (x - x2) ....(x - xn-1) /
(xn - x0) (xn - x1) (xn - x2) ....(xn - xn-1)) yn
Numerator → difference of x with all terms except xi
Denominator → difference of all terms xj from xi.
∏nj=0 (x - xj) (j ≠ i)
n
y= n yi
i=0 ∏j=0
(xi - xj) (j ≠ i)

Lagrange ' s Interpolation law can be

writter for inverse interpolation since it
iss merely a relation between two variables.
n ∏nj=0 (y - yj) (j ≠ i)
x= yi
i=0 ∏nj=0 (yi - yj) (j ≠ i)

Chapter 2
CURVE FITTING
 2019_Interpolation.nb 15

Let us have have n data of points. We

are interested to fit the data to a polynomial,
y = a0 + a1 * x + a2 * x2 + ... + an * xn
Depending upon different values of the parameters a0, a1 ..
an, we get different curves.
For each set of parameter values, we can find
the error, the difference of the point
from the theoretical value, for each point.
Error = (yt,i - yi)
n
Sum of of errors = e =  (yt,i - yi)
i=1
n
Sum of squares of errors = E =  (yt,i - yi)2
i=1

For an nth degree polynomial yt,i =

a0 + a1 * xi + a2 * xi2 + ... + an * xin
n
2
E =  a0 + a1 * xi + a2 * xi2 + ... + an * xin - yi
i=1

For finding the best fit we have to minimise the E with respect
to all the parameters a0, a1, ... an simultaneously.
∂ E ∂ E ∂ E ∂ E
= 0 & = 0 & ... .... &
∂ a0 ∂ a1 ∂ a2 ∂ an
n
∂ E ∂ 2 n 2
=  a0 + a1 * xi + a2 * xi + ... + an * xi - yi =
∂ a0 ∂ a0 i=1
n
2  a0 + a1 * xi + a2 * xi2 + ... + an * xin - yi = 0
i=1
n
∂ E ∂ 2 n 2
=  a0 + a1 * xi + a2 * xi + ... + an * xi - yi =
∂ a1 ∂ a1 i=1
n
2  a0 + a1 * xi + a2 * xi2 + ... + an * xin - yi xi = 0
i=1

... ... ... ... ... ... ... ...

16 2019_Interpolation.nb

n
∂ E ∂ 2 n 2
=  a0 + a1 * xi + a2 * xi + ... + an * xi - yi =
∂ an ∂ an i=1
n
2  a0 + a1 * xi + a2 * xi2 + ... + an * xin - yi xin = 0
i=1

∑1 ∑ xi ∑ xi 2 . ∑ xin-1 a0
∑ xi ∑ xi 2 ∑ xi 3 . ∑ xi n a1
∑ xi 2 ∑ xi 3 ∑ xi 4 . ∑ xin+1 a2
. =
. . . . .
. . . . . .
∑ xi n ∑ xin+1 ∑ xi 2 . ∑ xi2 n-1 an

∑ yi
∑ xi yi
∑ xi 2 yi
.
∑ xin-1 yi
First Degree Polynomial (Fitting to a straight line)
n
E =  (a0 + a1 * xi - yi)2
i=1

∂ E ∂ E
= 0 & = 0 →
∂ a0 ∂ a1
n n
∂ E ∂ 2
=  (a0 + a1 * xi - yi) = 2  (a0 + a1 * xi - yi) = 0
∂ a0 ∂ a0 i=1 i=1
n n
∂ E ∂ 2
=  (a0 + a1 * xi - yi) = 2  (a0 + a1 * xi - yi) * xi = 0
∂ a1 ∂ a1 i=1 i=1

1 ∑ xi a ∑ yi
  0 =  
∑ xi ∑ xi 2 a1 ∑ xi yi
Second Degree Polynomial (Fitting to a quadratic)

y = a0 + a 1 * xi + a 2 * xi 2
n
2
E =  a0 + a1 * xi + a2 * xi2 - yi
i=1
 2019_Interpolation.nb 17

∂ E ∂ E ∂ E
= 0 & = 0 & =0 →
∂ a0 ∂ a1 ∂ a2
∂ E
=
∂ a0
n n
∂ 2 2
 a0 + a1 * xi + a2 * xi - yi = 2  a0 + a1 * xi + a2 * xi2 - yi = 0
∂ a0 i=1 i=1
n
∂ E ∂ 2 2
=  a0 + a1 * xi + a2 * xi - yi =
∂ a1 ∂ a1 i=1
n
2  a0 + a1 * xi + a2 * xi2 - yi * xi = 0
i=1
n
∂ E ∂ 2 2
=  a0 + a1 * xi + a2 * xi - yi =
∂ a2 ∂ a2 i=1
n
2  a0 + a1 * xi + a2 * xi2 - yi * xi2 = 0
i=1

∑ 1 ∑ xi ∑ xi 2 a0 ∑ yi
∑ xi ∑ xi 2 ∑ xi 3 a1 = ∑ xi yi
∑ xi 2 ∑ xi 3 ∑ xi 4 a2 ∑ xi 2 yi
Equations which can be converted into polynomial equations
y = a0 exp (a1 x + a2 x2 + ....)
these equations can be converted into polynomial
equations by taking the loge on both sides.
log y = log a0 + a1 x + a2 x2 + ....
x+a2 x2 +....
Simillarly y = a0 10a 1
can be converted to
log10 y = log 10 a0 + a1 x + a2 x2 + ....