You are on page 1of 21

Summary B.V.

Ramana October 13, 2006 12:9


Chapter Summaries
Chapter-1
Vector Algebra, Theory of Equations, and
Complex Numbers
Vector algebra including dot product, cross product,
scalar triple product, vector triple product and their
application to work done, moment, volume of paral-
lelopiped are considered.
In the theory of equations, Cardons method of
obtaining algebraic solution of cubic equation and
Ferraris method of obtaining general solution of a
quartic equation are presented.
Algebra of complex numbers, De Moivres the-
orem and related properties are discussed. Method
of obtaining roots of a complex polynomial is pre-
sented.
Chapter-2
Differential Calculus
The nth derivatives of some elementary functions are
presented. Leibnitzs rule is used to nd derivative of
a product u(x) v(x).
For a function f continuous in [a, b] and differen-
tiable in(a, b) andf (b) = f (a) thenRolles theorem
proves that a c in (a, b) exists where f

(c) = 0. La-
granges Mean Value theorem is a generalization of
Rolles theorem when f (b) = f (a). It shows that a
c in (a, b) exists where f

(c) =
f (b)f (a)
ba
. Appli-
cation of LMV theorem to sign of derivative helps
to nd out whether f is increasing or decreasing.
Cauchys Mean Value theorem is a further general-
ization of LMV theorem. Finally Generalized Mean
Value theorem (Taylors theorem) prove that
f (a +h) =f (a) +hf

(a) +
h
2
2!
f

(a) +
+
h
n1
(n 1)!
)f
(n1)
(a)
+
h
n
(1 )
np
(n 1)!p
f
(n)
(a+h)
where 0 < < 1. It calculates f at a +h in terms
of f and its derivatives at a.
A Taylors series expansion of a differentiable
function f (x) about a point x = a is an innite se-
ries expressing f (x) in terms of f (a) and deriva-
tives of f (x) at a. When the centre of expansion is
a = 0, Taylor series reduces to Maclaurins series.
LHospitals rule enables to evaluate indeterminate
forms which are of the form
a
0
or

.
Curvature which measures the degree of bentness
is studied for curves in cartesian, polar and paramet-
ric form. The centre, radius and circle of curvature
for all the three forms is considered.
Evolute is the locus of the centres of circle of
curvature.
Envelope for a one-parameter family of curves
f (x, y, ) = 0 is obtained by eliminating fromthe
equation using f

(x, y, ) = 0.
Chapter-3
Partial Differentiation
Functions of several variables is very important in
Engineering Mathematics.
Limit of a function f (x, y) as (x, y) approaches
(a, b) exist if the same limit value is attained along
any path from (x, y) to (a, b).
f (x, y) is continuous at a point (a, b) if the limit
of f as (x, y) (a, b) equals to the value of the
function at (a, b).
Partial derivative w.r.t. a variable is obtained by
usual differentiation, holding the other remaining
1.1
Summary B.V.Ramana October 13, 2006 12:9
1.2 SUMMARY
variables constant and higher order derivatives by
successive differentiation.
For functions of two or more variables f (x, y, z)
where x, y, z are functions of a single variable t , then
the total derivative of f w.r.t. t is given by chain rule
for partial differentiation
df
dt
=
f
x
dx
dt
+
f
y
dy
dt
+
f
z
dz
dt
For change of variables, composite functions are
differentiated by chain rule
f
t
=
f
x
x
t
+
f
y
y
t
+
f
z
z
t
f
s
=
f
x
x
s
+
f
y
y
s
+
f
z
z
s
where f = f (x, y, z) and x = x(t, s),
y = y(t, s), z = z(t, s).
Given an equation f (x, y) = 0 where f (x, y) is
an implicit function, derivatives can be calculated
without explicitly solving y interms of x by
dy
dx
=
f
x
f
y
, f
y
= 0
d
2
y
dx
2
=
[f
xx
(f
y
)
2
2f
xy
f
x
f
y
+f
yy
(f
x
)
2
]
(f
y
)
3
, f
y
= 0
This can be extended to implicit equation of three
variables:
f (x, y, z) =0
y
x
=
f
x
f
y
, f
y
= 0
z
x
=
f
x
f
z
, f
z
= 0.
Eulers theoremfor homogeneous functions states
that
x
f
x
+y
f
y
= nf
and is useful in evaluating the L.H.S. of the above
equation as nf , without the necessity for calculation
of f
x
, f
y
.
Jacobian of two functions u(x, y) and v(x, y)
which are functions of two variables x and y is a
functional determinant.
J
_
u, v
x, y
_
=
(u, v)
(x, y)
=

u
x
u
y
v
x
v
y

For the inverse transformation


x = (u, v) and y = (u, v)
the Jacobian
J

=
(x, y)
(u, v)
having the property that J J

= 1.
Chain rule for Jacobians
(u, v)
(x, y)
=
(u, v)
(r, s)
(r, s)
(x, y)
is useful in change (transformation) of variables from
one coordinate system to the other.
Necessary condition for functional dependence of
two functions u = f (x, y) and v = g(x, y) is that
their Jacobian
J
_
u, v
x, y
_
= 0
in which case u can be expressed in terms of v (and
vice versa) through a functional relation between u
and v.
Error z for a function z = f (x, y) can be calcu-
lated approximately using differential d
z
.
Leibnitzs rule enables to differentiate under inte-
gral sign as
d
d
_
b
a
f (x, )dx =
_
b
a
f (x, )

dx
i.e., the order of differentiation and integration can
be interchanged.
Chapter-4
Maxima and Minima
By the use of Taylors theorem for functions of two
variables, f (a +h, y +k) canbe expandedinTaylor
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.3
series in powers of h and k as
f (a +h, y +k) =f (a, b) +[hf
x
(a, b) +kf
y
(a, b)]
+
1
2!
_
h
2
f
xx
(a, b) +2hkf
xy
(a, b)
+ k
2
f
yy
(a, b)
_
+ .
Alternatively Taylors expansion can be written in
powers of (x a) and (y b) as
f (x, y) =f (a, b) +[(x a)f
x
(a, b)
+(y b)f
y
(a, b)] +
1
2!
_
(x a)
2
f
xx
(a, b)
+ 2(x a)(y b)f
xy
(a, b) +(y b)
2
f
yy
(a, b)
_
+
1
3!
_
(x a)
3
f
xxx
(a, b)
+3(x a)
2
(y b)f
xxy
(a, b)
+3(x a)(y b)
2
f
yyx
(a, b)
+ (y b)
3
f
yyy
(a, b)
_
+
Essentially, the value of f at (x, y) is expressed in
terms of value of f and its derivatives evaluated at
(a, b).
Maclaurins expansionis Taylors expansionabout
origin (a = 0, b = 0) in powers of x and y.
The necessary and sufcient conditions for max-
ima and minima of a function of two variables is
derived from the Taylors expansion.
The stationary points are solutions of
f
x
= 0 and f
y
= 0
with r = f
xx
, s = f
xy
, t = f
yy
the sufcient condi-
tions are:
Maximum when rt s
2
> 0 and r < 0
Minimum when rt s
2
> 0 and r > 0
Saddle point when rt s
2
< 0.
Lagranges multipliers method determines the
stationary points of a constrained extrema problem,
where f (x, y, z) is to be extremized subject to the
constraint g(x, y, z) = c, by introducing the aux-
iliary function F(x, y, z) = f (x, y, z) +g(x, y, z)
and solving
F
x
= 0, F
y
= 0, F
z
= 0
and the constraint g(x, y, z) = c.
Although this method can easily be extended to any
number of variables and any number of constraints,
the main disadvantage is it can not determine the
nature of the stationary points.
Chapter-5
Curve Tracing
Tracing of curves whose equations are in the carte-
sian form, polar form and parametric form is consid-
ered.
For cartesian curves, symmetry about the axes, or
origin or y = x or y = x simplies curve tracing.
Determine the imaginary region in which curve does
not exist. Vertical or horizontal or oblique asymp-
totes are next studied. If origin lies on curve, tangents
at origin and other intercept points are determined.
Nature of multiple point node, cusp or isolated point
is found by the number of tangents at that point. Sign
of the derivative gives the nature of the curve whether
it is increasing or decreasing or stationary where ex-
trema occurs. Several important cartesian curves are
traced for reference.
In tracing polar curves identify the symmetry of
the curve about the initial line or line through pole
and perpendicular to initial line or = /4 or =
3/4. Region of existence is found. Similar as in
the cartesian curves, status of pole, points of inter-
section, tangents thereat and sign of derivative are
useful. Particularly calculate values of r for several
values of . Parametric equations are handeled in a
similar way.
Chapter-6
Integral Calculus
Reduction formulae, which reduces a given integral
to a known integration form by repeated application
of integration by parts, are presented.
Method of determining plane area bounded by a
given set of plane curves in cartesian, parametric and
polar form is presented.
Summary B.V.Ramana October 13, 2006 12:9
1.4 SUMMARY
Length of the arc of a plane curve is studied. When
the equation of the curve is in the cartesian from
y = f (x) the length of the curve S is given by
S =
_
b
a
_
1 +
_
dy
dx
_
2
dx
For curve in parametric from x = x(t ), y = y(t )
S =
_
t
2
t
1
_
_
dx
dt
_
2
+
_
dy
dt
_
2
dt
In polar form r = r()
S =
_

2

1
_
r
2
+
_
dr
d
_
2
d
Volume V of a solid of revolution generated by
revolving a plane area R about say x-axis can be
obtained by cylindrical disk method as
V =
_
b
a
y
2
dx
About y-axis, V =
_
d
c
x
2
dy.
When axis of rotation is not part of the area R then
cylindrical shell method is useful. For rotation about
x-axis
V =
_
d
y=c
(2y)(x)dy
About y-axis, V =
_
b
x=a
(2x)y dx
When the equation of the curve binding the area R
is in parametric form, then the above results are con-
verted to t
V =
_
t
2
t
1
y
2
dx
dt
dt. etc.
For polar curves with rotation about the initial line
= 0,
V =
_

2

1
2
3
r
3
sin d
If the rotation is about the line =

2
, then
V =
_

2

1
2
3
r
3
cos d
The surface area of a solid of revolution generated
by revolving a plane curve about say x-axis in the
cartesian form is
S =
_
b
x=a
2y
_
1 +
_
dy
dx
_
2
dx
In parametric form, about x-axis
S =
_
t
2
t
1
2 x(t )
_
_
dx
dt
_
2
+
_
dy
dt
_
2
dt
In polar form about = 0
S =
_

2

1
2(r sin )
_
r
2
+
_
dr
d
_
2
d
and about the line =

2
.
S =
_

2

1
2(r cos )
_
r
2
+
_
dr
d
_
2
d.
Improper integrals of three kinds are considered
which are useful in nding the area of an unbounded
region.
Chapter-7
Multiple Integrals
1. Multiple integrals, double integrals and triple in-
tegrals are considered.
2. Double integral is evaluated as two-fold iterated
integral.
3. Application of double integral to nd the area,
mass, centroid and moments of inertia of a plane
region are studied. Double integral is also used to
evaluate the volume of a solid region.
4. The evaluation of double integral, in some cases,
becomes much easier with the change of order of
integration.
5. With the change of variables fromone coordinate
system to the other, in several cases, the double
integral can be evaluated in a simpler way.
6. Applying the change of variables, double inte-
grals in polar coordinates are used to nd area,
mass, centroid and moments.
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.5
7. Triple integral is evaluated as three fold iterated
integral and is used to nd the volume, mass, cen-
troid and moments of inertia of a solid region.
8. The general change of variables in triple inte-
gral is considered. The special cases of change
of variables from cartesian to cylindrical and
from cartesian to spherical coordinates were also
discussed.
Chapter-8
Ordinary Differential Equations: First Order
and First Degree
Mathematical modelling describes a natural process
or a physically system in mathematical terms yield-
ing reproduciable results which can be used for pre-
diction.
1. First order and rst degree ordinary differential
equations of the form
dy
dx
= f (x, y) (1)
are studied.
2. Methods for obtaining general solutions y =
(x, c), involving one arbitrary constant of (1)
are presented in the following cases.
A. D.E. is separable in which case the variables
x and y are separated and D.E. takes the
form
F(x)dx +G(y)dy = 0
which on integration yields the general so-
lution.
B. By a substitution y = ux, a homogeneous
D.E. becomes separable and can be solved
by method A.
C. With the transformation x = x
1
+h, y =
y
1
+k (shift of origin) a non-homogeneous
D.E. of the form
dy
dx
=
a
1
x +b
1
y +c
1
a
2
x +b
2
y +c
2
reduces to homogeneous D.E. discussed
in B.
D. A D.E. Mdx +Ndy = 0 is an exact D.E. if
M
y
=
N
x
in which case there exists a function f (x, y)
such that
df = Mdx +Ndy = 0
resulting on integration the solution
f (x, y) = c.
E. When a D.E. is non exact, it can be made
an exact D.E. by multiplying it by the in-
tegrating factor (I.F.) f (x, y). Methods of
obtaining I.F. F(x, y):
i. By inspection (regrouping terms appropri-
ately) [see table in Section 8.8 for ready ref-
erence].
ii. When
_
M
y
N
x
_
/N = g(x) is a function
of x alone, then
I.F. = e
_
g(x)dx
.
iii. When
_
N
x
M
y
_
/N = h(y) is a function
of y alone, then
I.F. = e
_
h(y)dy
.
iv. When D.E. is homogeneous. Then the I.F.
is
(xM +yN)
1
.
v. When D.E. is of the form
y g(xy)dx +x h(xy)dy = 0
Then the I.F. is
(xM yN)
1
.
vi. When D.E. is of form
x
a
y
b
(my dx +nx dy) +x
c
y
d
(py dx +qx dy) = 0
Then the I.F. is x
h
y
k
.
3. A rst order linear D.E.
dy
dx
+p(x)y = Q(x)
has an I.F. = e
_
p(x)dx
and the solution is
y e
_
p(x)dx
=
_
e
_
p(x)dx
Q(x)dx
Summary B.V.Ramana October 13, 2006 12:9
1.6 SUMMARY
4. A Bernoulli equation (which is nonlinear)
dy
dx
+P(x)y = Q(x)y
a
can be transformed to a linear D.E. by the sub-
stitution
z = y
1a
5. A D.E. is formed (or obtained) by eliminating
the arbitrary constants fromthe general solution
and its derivative.
6. Application of D.E. to geometry is considered
to nd the equations and lengths of tangent, nor-
mal, subtangent and subnormal in cartesian and
polar coordinates.
A. Given a one-parameter family of curves
F(x, y, c) = 0 (2)
and its D.E. (obtained by elimination of c
from (2)
dy
dx
= f (x, y)
The orthogonal trajectories (the second one-
parameter family of curves) are given by the
general solution of the corresponding DE
dy
dx
=
1
f (x, y)
B. In polar coordinates, for given equation r =
f () and its D.E.
r d
dr
=
rP(r, )
Q(r, )
The orthogonal trajectories are given by the
general solution of the corresponding D.E.
r d
dr
=
Q
rP
7. A simple application of rst order D.E. is New-
tons law of cooling described by
dT
dt
= k(T T
A
)
with general solution given by
T (t ) = T
A
+c e
kt
(3)
The arbitrary constant c and the proportionality
constant k in (3) are determined using the given
two initial conditions.
8. Application of D.E. to the lawof natural growth
(such as growth of bacteria) and natural decay
(like radiumdecomposition) is also considered.
9. The velocity v of a body to escape from earth
is modeled by rst order DE
v
dv
dr
=
kM
r
2
with solution
v
2
2
=
kM
r
+c. The minimum ve-
locity of escape is v
0
=
_
2kM
R
.
10. The velocity current (or charge) in a simple
electric curcuit is obtained as solution of rst
order DE
dI
dt
+
R
L
I =
E(t )
L
for RL-circuit
dI
dt
+
1
RC
I =
1
R
dE
dt
for RC-circuit.
Chapter-9
Linear Differential Equations of Second
Order and Higher Order
Linear differential equations of second order
and higher order (both homogeneous and non-
homogeneous) are considered.
1. A system of functions is linearly independent or
dependent according as its Wronskian is non zero
or zero.
2. The general solution of a homogeneous second
order linear D.E. with constant coefcients is
obtained depending on the nature of roots of its
corresponding auxiliary equation.
3. The superposition or linearity principle is appli-
cable to only homogeneous D.E.
4. The general solution of a non-homogeneous
linear D.E. with constant coefcients is the
sum of the complimentary function (which is
the solution of the corresponding homogeneous
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.7
equation) and the particular integral (which is a
particular solutionof the givennon-homogeneous
D.E.) i.e.,
G.S. = C.F +P.I
5. Using the inverse operator, methods of obtaining
particular integral, when the R.H.S. function
F(x) of the non-homogeneous D.E. is of the
form e
ax
, cos(ax +b), sin(ax +b), x
m
, e
ax
V(x)
and xV(x) are studied.
6. Cauchy-Euler equationandLegendre linear equa-
tion are D.E. with variable coefcients but can be
reduced to D.E. with constant coefcients by sub-
stitutions of x = e
t
and ax +b = e
t
respectively.
7. Method of variation of parameters is a powerful
method to solve any linear non-homogeneous
D.E. (even with variable coefcients) wherein
the parameters (constants) in the given (known)
complimentary function are varied to obtain a
particular integral.
8. Method of elimination is used of solve a system
of simultaneous linear differential equations
with constant coefcients.
9. Method of undetermined coefcients is another
useful method of obtaining P.I. of an nth order
linear non homogeneous DE with constant co-
efcients when the RHS function is exponential,
polynomial, sine, cosine or sum and products of
such functions.
10. Method of reduction of order enables to nd a
second linearly independent solution y
2
(x) of
a second order linear homogeneous DE with
constant coefcients when the rst solution y
1
(x)
is known.
11. Certain second and higher order DEwith variable
coefcients can be reduced to lower order DE by
certain transformations and can be solved by the
earlier methods.
12. Applications of second order DE:
(a) Simple harmonic motion is described by
d
2
x
dt
2
=
2
x
with displacement (solution) x given by
x(t ) = b cos t
(b) The free and forced, damped and undamped
oscillations of a mass spring mechanical system
are described by
m x(t ) +cx
.
(t ) +kx(t ) = F(t )
The resonance phenomenon in forced
undamped oscillations is also considered.
(c) The electric current I (and charge Q) in an
RLC-circuit is obtained as solution of
L
d
2
I
dt
2
+R
dI
dt
+
1
c
I =
dE
dt
_
or L
d
2
Q
dt
2
+R
dQ
dt
+
1
c
Q = E(t )
_
(d) The equation of angular motion (t ) of the
forced damped simple pendulum with driving
force F(t ) and variable length L(t ) is given by
mL

+(2mL
.
+cL)
.
+mg sin = F(t )
The linearized pendulum equation with
xed length L is described by
mL

+cL
.
+mg = F(t )
Chapter-10
Series Solutions
The solution to DE with variable coefcients y

+
P(x)y

+Q(x)y = 0 can be obtained in the form of


an innite power series.
(a) When x
0
is a regular point of DEthen power series
method is used in which a solution of the form

n=0
c
n
(x x
0
)
n
is assumed. The unknown coefcients c
n
are deter-
mined from recurrence relations.
(b) When x
0
is a regular singular point of DE. Then
in Frobenius (Generalized Power Series) method a
solution of the form
(x x
0
)
r

n=0
c
n
(x x
0
)
n
is assumed. Solutions are obtained depending on the
nature of roots r of the indicial equation.
Summary B.V.Ramana October 13, 2006 12:9
1.8 SUMMARY
A piecewise function f (x) can be represented as
an innite series known as Generalized Fourier series
in terms of an orthogonal set of functions.
Gram-Schmidt orthogonalization process is used
to an orthonormal basis from a given set of non-
orthogonal basis of vectors. The eigen values and
eigen functions of a regular, singular and periodic
Sturm-Liouville problem are considered.
Chapter-11
Special FunctionsGamma, Beta, Bessel
and Legendre
1. Multiple integrals, double integrals and triple in-
tegrals are considered.
2. Double integrals is evaluated as two-fold iterated
integral.
3. Application of double integral to nd the area,
mass, centroid and moments of inertia of a plane
region are studied. Double integral is also used to
evaluate the volume of a solid region.
4. The evaluation of double integral, in some cases,
becomes much easier with the change of order of
integration.
5. With the change of variables fromone coordinate
system to the other, in several cases, the double
integral can be evaluated in a simpler way.
6. Applying the change of variables, double inte-
grals in polar coordinates are used to nd the vol-
ume, mass centroid and moments.
7. Triple integral is evaluated as three fold iterated
integral and is used to nd the colume, mass cen-
troid and moments of inertia of a solid region.
8. The general change of variables in triple integral
is considered. The special cases of change of vari-
ables fromcartesian to cylindrical and fromcarte-
sian to spherical coordinates were also discussed.
Chapter-12
Laplace Transform
Laplace Transform is a very powerful technique to
solve linear differential equations both ordinary and
partial and systemof simultaneous differential equa-
tions. The Laplace transformof a given function f (t )
is dened as
L{f (t )} =
_

0
e
st
f (t )dt = F(s) (1)
The crucial idea in L.T. is that (1) replaces opera-
tions of calculus by operations of algebra and has the
added advantage that the initial conditions are taken
care.
Although L.T. of several functions can be derived
from (1), in practice L.T. are obtained by applying
some important properties such as:
1. L.T. operator is linear.
2. First shifting: Multiplication by e
at
amounts to
replacement of s by s a.
3. 2nd shifting: replacement of t by t a in f (t )
amounts to multiplication of F(s) by e
as
.
4. L.T. of derivative amounts to multiplication of
F(s) by s (approximately).
5. L.T. of integral amounts to division by s.
6. Multiplication of f (t ) by t has the effect of dif-
ferentiating F(s).
7. Division of f (t ) by t has the effect of integration
of F(s).
8. L.T. of a periodic function of period p is obtained
by integrating e
st
f (t ) in the interval (o, p), mul-
tiplied by the factor (1 e
ps
)
1
.
Thus the application of L.T. to a O.D.E. or
a system of D.E., with given I.C.(s) reduces to
algebraic equation (s) whose inverse Laplace
transform(s) can be obtained from:
i. L.T. tables.
ii. Use of important properties of L.T.
iii. Use of partial Fractions
iv. Use of convolution Theorem.
The importance of convolution theoremcomes
from the fact that the I.L.T of H(s), which is writ-
ten as the product of two functions F(s) and G(s)
whose I.L.T. are known, is given by the convolu-
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.9
tion of f (t ) and g(t ) dened as
L
1
{H(s)} = f g =
_
t
0
f (u)g(t u)du
A list of some important properties and a gen-
eral list of L.T. are presented in sections 12.12
and 12.13.
Chapter-13
Matrices
A brief introduction of matrices is presented. Matrix
is a rectangular array of mn numbers. Square ma-
trix when m = n i.e., the number of rows equals to
number of columns.
Inverse of a matrix Aexists only if Ais square and
non-singular (|A| = 0). Then A
1
A = A
1
A = I.
Inverse is unique.
(AB)
1
=B
1
A
1
and
(A
1
)
T
=(A
T
)
1
A
1
=adj A/|A|
where adjA = [A
ij
]
T
with A
ij
as the cofactors of a
ij
of A.
Inverse of a matrix can also be obtained by Gauss-
Jordan method with the application of elementary
row operations.
Rank of a matrix is the largest order of a non-zero
minor.
Rank of matrix is the number of non-zero rows
in the normal form or echelon form. A system of m
equations in n unknowns
AX = B
is consistent if rank of A and rank of the augmented
matrix is same. Further if r(A) = r(A|B) n then
a unique solution exists.
If r =r(A) =r(A|B) < n then innitely many so-
lutions exists in terms of (n r) arbitrary variables.
If r(A) = r(A|B), system is inconsistent and has
no solution at all.
When m = n, the system can be solved by
Cramers rule x
i
=
D
i
D
for i = 1 to n and also by ma-
trix inversion X = A
1
B. Gaussian elimination and
Gauss-Jordan elimination methods reduce a system
of m non-homogeneous equations in n unknowns
into an upper triangular and diagonal systems respec-
tively.
For homogeneous systemof equations, non-trivial
solutions exist if m < n. If m = n, non-trivial solu-
tions exist if its coefcient determinant is zero.
The solution of a nonhomogeneous system of
n equations in n unknowns is obtained by LU-
decomposition in which the coefcient matrix A is
factored into a lower triangular matrix L and an up-
per triangular matrix U. Gaussian elimination is also
used for LU-decomposition. Solution to tridiagonal
system by Crouts reduction is presented.
Chapter-14
Eigen Values and Eigen Vectors
Only Square Matrices are Considered
Eigen values, eigen vectors and their properties for
both real and complex matrices are studied. Eigen
(characteristic or latent) value of a matrixAis a scalar
such that
AX = X
for an X = 0. Here X is known as the eigen vec-
tor corresponding to eigen value . Eigen values of
A are determined from the characteristic equation
|A I| = 0. The corresponding eigen vectors are
obtained by solving (A I)X = 0 for a specic .
For an n-square matrix, at least one and at most n
eigen values exist.
Inverse of a matrix A can also be obtained using
CayleyHamilton theorem which states that every
square matrix satises its own characteristic polyno-
mial. An n-square matrix A can be diagonalized as
D = P
1
AP
if there exist n linearly independent eigen vectors
of A. Here P is known as modal matrix which
contains the n linearly independent eigen vectors as
its n columns. Powers of a matrix A, then can be
calculated as
A
n
= PD
n
P
1
The eigen values of a Hermitian, (A = A
T
) (real
symmetric A = A
T
) matrix are real; of a Skew-
Hermitian (A = A
T
) (real Skew-symmetric A =
A
T
) matrix are purely imaginary or zero and of
Summary B.V.Ramana October 13, 2006 12:9
1.10 SUMMARY
a unitary (A
T
= A
1
) (real orthogonal A
T
= A
1
)
matrix are of absolute value 1.
The column (and row) vectors of a unitary (orthog-
onal) matrix form an unitary (orthonormal) system
of vector. By an orthogonal transformation, a real
quadratic form Q = X
T
AX can be transformed (re-
duced) to canonical form Y
T
DY where D is the di-
agonal matrix with eigen values of Aas the diagonal
elements.
A quadratic form is classied as positive denite
(semi-denite), negative denite (semi-denite) or
indenite according as the eigen values are all posi-
tive (includes one zero), negative or otherwise.
Chapter-15
Vector Differential Calculus: Gradient,
Divergence and Curl
Vector differential calculus extends the differential
calculus of scalar functions to the vector functions,
with all the basic rules and formulas quite similar.
Derivative of a vector function in component form
is obtained by taking derivatives of the components
i.e., if
F = F
1
i +F
2
j +F
3
k
Then
dF
du
=
dF
1
du
i +
dF
2
du
j +
dF
3
du
k
A similar denition for partial derivative is
obtained by taking the partial derivatives of the
components with respect to the corresponding
independent variable
F
u
=
F
1
u
i +
F
2
u
j +
F
3
u
k
The del or nabla operator is
= i

x
+j

y
+k

z
Gradient of a scalar function f is a vector
grad f = f = i
f
x
+j
f
y
+k
F
z
= vector
gives the direction of maximum increase of f .
f is the normal to the surface f (x, y, z) = c.
Directional derivative of f at point p in the direction
of a vector a is
df
ds
= f
a
|a|
Divergence of a vector function A is a scalar
divA = A =
A
1
x
+
A
2
y
+
A
3
z
= scalar
When A = 0, the vector eld A is solenoidal.
Curl of a vector function A is a vector
curlA = A =

i j k

z
A
1
A
2
A
3

= vector
Related properties of sums, differences and
products of gradient, divergence and curl are
quite similar to the differential calculus properties,
keeping the order intact.
In the second-order differential operators

2
f = f =

2
f
x
2
+

2
f
y
2
+

2
f
z
2
and
curl (f ) =0
div (curl A) =0.
Expressions for grad, div, curl and Laplacian in
curvilinear, cylindrical and spherical coordinates are
presented in the Table 15.1 for convenience.
Chapter-16
Vector Integral Calculus
In vector integral calculus the results of (ordinary)
integral calculus are extended to vector functions.
1. The indenite integral of a vector function f (u) in
the component formis the sumof three (ordinary)
integrals of its scalar components i.e., if
f (u) = f
1
(u)i +f
2
(u)j +f
3
(u)k
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.11
then
_
f (u)du =i
_
f
1
(u)du +j
_
f
2
(u)du
+k
_
f
3
(u)du
similarly, denite integral froma to b is given by
_
b
a
f (u)du = F(b) F(a).
2. Line integral is a generalization of an ordinary
integral. In a line integral the integrand is inte-
grated along the curve.
Line integral is used to calculate
work done by a force:
_
F dr
area of a plane region: A =
1
2
_
x dy y dx
scalar potential of
a conservative eld F:
_
P
2
P
1
F dr =
_
P
2
P
1
d
= (P
2
) (P
1
)
3. Surface integral is a generalization of a double
integral. In a surface integral the integrand is
integrated along a curved surface.
Surface integral is used to compute the surface
area of a curved surface
S =
_ _
R
Fx
2
+Fy
2
+Fz
2
|F
z
|
dxdy
The ux of a vector function F across a surface
S is given by the surface integral
_ _
F ndS
4. Greens theoreminplane transforms line integrals
to double integrals and vice versa.
_
c
Mdx +Ndy =
_
R
_ _
N
x

M
y
_
dxdy
Area A of a plane region bounded by a simple
closed curve c can be calculated using Greens
theorem as
A =
1
2
_
c
xdy ydx in cartesian coordinates
A =
1
2
_
c
r
2
d in polar coordinates
5. Stokes theorem is a generalization of the
Greens theorem, and connects line integral with
surface integral.
__
S
( A) nds =
_
c
A d r
Circulation of A around a closed curve c is
_
c
A dr
6. Gauss divergence theorem is a transformation
from surface integral to volume integral and vice
versa.
__

S
A n ds =
___
V
Adv
Chapter-17
Fourier Series
Fourier series, which is an innite trigonometric
series representation of a periodic function f (x) in
the interval (, +2) is
f (x) =
a
0
2
+

n=1
(a
n
cos nx +b
n
sin nx)
where the Fourier coefcients are
a
0
=
1

_
+2

f (x) dx
a
n
=
1

_
+2

f (x) cos nx dx, for n = 1, 2, 3 . . .


b
n
=
1

_
+2

f (x) sin nx dx, for n = 1, 2, 3, . . .


For function f (x) which is even in the interval
(, ), the Fourier series reduces to Fourier cosine
series given by
f (x) =
a
0
2
+

n=1
a
n
cos nx
where
a
0
=
2

_

0
f (x) dx
and
Summary B.V.Ramana October 13, 2006 12:9
1.12 SUMMARY
a
n
=
2

_

0
f (x) cos nx dx, for n = 1, 2, 3
For odd function f (x) in the interval (, )
Fourier series reduces to Fourier sine series
f (x) =

n=1
b
n
sin nx
where
b
n
=
2

_

0
f (x) sin nx dx
Fourier series can also be expanded in an interval
(c, c +2L) of period 2L as
f (x) =
a
0
2
+

n=1
_
a
n
cos
nx
L
+b
n
cos
nx
L
_
where
a
0
=
1
L
_
c+2L
c
f (x) dx
a
n
=
1
L
_
c+2L
c
f (x) cos
nx
L
dx
b
n
=
1
L
_
c+2L
c
f (x) sin
nx
L
dx
For f (x) which is not periodic and dened only in
(half) interval (0 < x < L) the half-range Fourier
cosine series are given by
f (x) =
a
0
2
+

a
n
cos
nx
L
with
a
0
=
2
L
_
L
0
f (x) dx, a
n
=
2
L
_
L
0
f (x) cos
nx
L
dx.
Half-range fourier sine series by
f (x) =

n=1
b
n
sin
nx
L
with
b
n
=
2
L
_
L
0
f (x) sin
nx
L
dx
For a function f (x) given in tabulated form, the
approximate values of the Fourier coefcients a
0
, a
m
,
b
m
are calculated in practical harmonic analysis.
Chapter-18
Partial Differential Equations
Partial Differential Equations (PDE) appear in prob-
lems where the number of independent variables is
two or more. Formation of PDE by elimination of ar-
bitrary constants and functions is considered. Meth-
ods of obtaining the general solution of Lagranges
linear equation
P(x, y)p +Q(x, y)q = R(x, y)
are presented. Complete solutions of non-linear rst
order PDE of types f (p, q) = 0, f (z, p, q) = 0,
f (x, p) = g(y, q) and Clairaut equation z = px +
qy +f (p, q) are obtained. Charpits method of ob-
taining the complete solution of any rst-order non-
linear PDE is presented. Methods of obtaining the
general solution of both homogeneous and non-
homogeneous PDEwithconstant coefcients is stud-
ied when the RHS function is exponential, sine, co-
sine, x
m
y
n
or e
ax+by
v(x, y). Cauchys PDE with
variable coefcients is also considered. Monges
method of obtaining the general solution of most
general non-linear PDE of second order is discussed.
Certain second order PDE whose solutions can be
obtained are also considered.
Chapter-19
Application of Partial Differential Equations
Separation of variables technique is a powerful
method of solving initial boundary value problems
involving PDE. It is assumed that the solution is sep-
arable, which reduces PDE to two ordinary DE. The
general solution is then obtained by using the bound-
ary and initial conditions and the linearity principle.
Using this technique, three very important second or-
der linear homogeneous PDE are solved, as follows.
(a) The parabolic one-dimensional heat equation
u
t
= a
2

2
u
x
2
is derived and solved by method of separa-
tion of variables for homogeneous (zero), non-
homogeneous (non-zero) and with one-end, both
ends insulated boundary conditions.
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.13
(b) The hyperbolic one-dimensional wave equation

2
y
t
2
= a
2

2
y
x
2
is derived and solved with given initial displace-
ment given initial velocity and both given dis-
placement and velocity, conditions.
(c) The elliptic two-dimensional Laplaces equa-
tion

2
u
x
2
+

2
u
y
2
= 0 is solved in a rectangle and
in a semi-innite strip with non-insulated and in-
sulated edge conditions.
(d) Laplaces equation in polar coordinates

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0
and Dirichlet circular disk problem are also
solved.
(e) Two-dimensional heat equation
u
t
= c
2
_

2
u
x
2
+

2
u
y
2
_
is derived and solved in a rectangle with non-
insulated and insulated edges using double
Fourier series.
(f) Two-dimensional wave equation

2
u
t
2
= c
2
_

2
u
x
2
+

2
u
y
2
_
is derived and solved for a rectangular membrane
involving double Fourier series.
(g) The vibrations of a circular membrane are mod-
eled by

2
u
t
2
= c
2
_

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
_
is solved involving Bessels equations.
(h) The transmission line equations

v
x
= Ri +L
i
t
and

i
x
= Gv +c
v
t
are derived and the radio equation and telegraph
equation are solved.
Chapter-20
Fourier Integral, Fourier Transforms and In-
tegral Transforms
Fourier integral representation of a non-periodic
function f (x) in < x < is a natural gener-
alization of the Fourier series representation (expan-
sion) of a periodic function f (x) in a nite interval
L < x < L.
Fourier integral of f (x) is
f (x) =
_

0
[A() cos(x) +B() sin(x)]dx
where A() =
1

f (t ) cos(t )dt
B() =
1

f (t ) sin(t )dt
or f (x) =
1

__

f (t ) cos((t x))dt
_
d
or in the complex (exponential) form
f (x) =
_

_
4
2
_

f (t )e
it
dt
_
e
ix
d
Fourier transform of f (x) is
F{f (x)} =
1
2
_

f (x)e
ix
dx = F()
Then the inverse Fourier transform of F() is
f (x) = F
1
{F(x)} =
_

F()e
ix
d
Fourier cosine transform of f (x) is
F
c
() =
_

0
f (x) cos x dx
Fourier sine transform of f (x) is
F
s
{f (x)} =
_

0
f (x) sin x dx
Fourier transform is useful in solving initial bound-
ary value problems such as the temperature in a semi-
innite bar, steady-state temperature in a semi in-
nite strip. In such problems the application of Fourier
transform reduces the one-dimensional heat PDE,
two dimensional Laplace PDEto ordinary DE. When
f (x) is dened in 0 < x < L, nite Fourier sine and
Summary B.V.Ramana October 13, 2006 12:9
1.14 SUMMARY
cosine transforms are used. Parsevals identity for
Fourier transforms is also considered.
Chapter-21
Linear Difference Equations and Z-
Transforms
Difference equations arise when differences in rela-
tions is considered. First and second order homoge-
neous difference equations with constant coefcients
are solved. Using method of undetermined coef-
cients and inverse operator method, general solution
of non-homogeneous linear difference equations is
obtained when the RHS is of the form a
n
, sin n,
cos n, n
m
, a
n
v(n).
The z-transform of a sequence u
n
is
z(u
n
) =

n=0
u
n
z
n
= U(z)
z-transformoperates ona sequence andis the discrete
analogue of Laplace transform. The damping rule,
shifting property. Initial and Final Value theorems,
Convolution theorems are useful in solving linear
differences equations.
Chapter-22
Complex Function Theory
A complex function f (z) of a complex variable z is
said to be continuous at z
0
if
lim f (z)
z z
0
= f (z
0
).
and differentiable at z
0
if
lim
z 0
f (z +z) f (z)
z
.
exists along any path as z 0. f (z) is analytic
at z
0
if it is differentiable at every point (including
z
0
) of some neighbourhood of z
0
. Analytic function
f (z) = u +iv satisfy Cauchy-Riemann equations
u
x
=
v
y
,
u
y
=
v
x
.
u and v satisfy Laplaces equation and are known as
harmonic functions and conjugate harmonic of each
other. The derivative can be evaluated using
f

(z) = u
x
+iv
x
= v
y
iu
y
Using Cauchy-Riemann equations, the real part
of an analytic function can be determined from the
given imaginary part (or vice versa).
Cauchy-Riemann equations in polar form are
u
r
=
1
r
v

,
v
r
=
1
r
u

Chapter-23
Complex Integration
Complex line integral of a complex function f (z)
from z
1
to z
2
along a curve c is dened by
_
c
f (z)dz =
_
z
2
z
1
(udx vdy) +i
_
z
2
z
1
vdx +udy
For f (z) analytic, the line integral is evaluated for
any path joining z
1
to z
2
by
_
z
2
z
1
f (z)dz = F(z
2
) F(z
1
)
where F

(z) = f (z).
For non-analytic f (z), the evaluation of line in-
tegral is done making use of the path c, z = z(t ),
a t b by
_
c
f (z)d
z
=
_
b
a
f (z(t ))
dz
dt
dt
The basic property of an analytic function f (z) is
established by Cauchys theorem which states that
_
c
f (z)dz = 0
for any simple closed curve c in the simply connected
domain of its analyticity. Cauchys integral formula
which is a consequence of Cauchys theorem estab-
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.15
lishes relation between value of analytic function at
interior point and boundary value of function by
_
c
f (z)
z z
0
dz = 2if (z
0
)
The generalized Cauchys integral formula
_
c
f (z)
(z z
0
)
n+1
dz =
2i
n!
f
(n)
(z
0
)
shows that for an analytic function, derivatives of all
orders exist and are analytic.
A power series of the form

n=0
a
n
(z z
0
)
n
rep-
resents an analytic function. Conversely every ana-
lytic function can be represented by a power series
knownas Taylors series; f (z)=

n=0
f
(n)
(z
0
)
n!
(zz
0
)
n
.
Maclaurins series is Taylors series about origin;
f (z) =

n=0
f
(n)
(0)
n!
z
n
.
An analytic function f (z) can be expanded about a
singular point z
0
in Laurent series consisting of both
positive and negative integer powers of z z
0
as
f (z) =

n=0
a
n
(z z
0
)
n
+

n=1
b
n
(z z
0
)
n
valid in the annulus region 0 < |z z
0
| < R.
The isolated singular point z
0
of f (z) is classied as
removable singularity, pole of order m or essential
singularity according as the number of terms in the
principal part

n=1
b
n
(zz
0
)
n
of the Laurent series expan-
sion of f (z) about z
0
is zero, nite (m) or innite.
z
0
is a kth order zero of f (z) if f, f

, f

, . . . , f
(k1)
the derivatives at z
0
are all zero and f
(k)
(z
0
) = 0.
Simple pole(or zero) is pole (or zero) of order one.
Chapter-24
Theory of Residues
Residue of f (z) at z = z
0
is the coefcient b
1
of
(z a)
1
in Laurent series expansion of f (z) about
z
0
and is
Res f (z)
z=z
0
= b
1
=
1
2i
_
c
f (z) dz
Residue theorem states that for an analytic func-
tion f (z),
_
c
f (z) dz = 2i

Res f (z)
where the summation is at all poles of f (z) which
are inside the simple closed curve c.
Residue theorem is useful in evaluating real de-
nite and improper integrals of the following types:
I.
_
2
0
F(sin , cos )d =
_
c
G(z) dz
= 2i

Res G(z)
where the summation is at all poles of G(z) inside
the unit circle c : |z| = 1.
II.
_

f (x) dx = 2i

Res f (z)
where the summation is at all poles of f (z) in the
upper half plane.
III.
_

f (x) cos mxdx = 2

Im.Res
_
f (z)e
imz
_
_

f (x) sin mxdx = 2

Re.Res
_
f (z)e
imz
_
where the summation is at all poles of f (z)e
imz
in
the upper half plane
IV. principal value of
_

f (x) dx
= 2i

Res f (z) +i

Resf (z)
where the rst summation extends overall poles
of f (z) in the upper half plane and the sec-
ond summation overall simple poles on the real
axis.
Argument principle states that if N and P are the
number of zeros and poles of an analytic function
f (z) within simple closed curve c then N P =
1
2

variation of argument of f (z) as z traverses c com-


pletelyonce.
Rouches theorem states that the number of zeros
of f (z) +g(z) and f (z) are same inside c if f, g are
analytic and |f | > |g| on c.
Fundamental theorem of algebra assures that
every polynomial of degree n has n zeros.
Summary B.V.Ramana October 13, 2006 12:9
1.16 SUMMARY
Liouvilles theorem states that an entire and
bounded function f (z) is constant.
Using argument principle, the number of zeros of
a complex polynomial each quadrant is determined.
Chapter-25
Conformal Mapping
Conformal mapping (or transformation) w = f (z)
preserves angles, both in magnitude and sense,
between any two curves and their images.
An analytic function is conformal everywhere
except at critical points where the derivative is zero.
Linear transformation w = az +b which is a
combination of rotation, scaling and translation is
conformal everywhere except for f

(z) = a = 0. It
transforms a circle in z-plane onto a circle in the
w-plane.
Inversion and reection transformation w =
1
z
is
an inversion w.r.t. a unit circle followed by reection
about the real axis. It is conformal everywhere except
at z = 0. It transforms circles onto circles.
w = z
n
for n integer > 1 is conformal everywhere
except at z = 0 and . It spreads an angular region
onto a half plane.
w = e
z
is conformal everywhere. It transforms a
rectangular stripof width ontothe upper half plane;
with the left semi-innite strip (x < 0) onto the in-
terior of a unit circle in the upper half plane and the
right semi-innite strip (x > 0) onto the exterior of
unit circle in the upper half plane.
Log function is considered as the inverse of the
exponential function.
w = sin z is conformal everywhere except at the
critical points z = (2n 1)

2
, n = 0, 1, 2, . . . It
transforms vertical and horizontal lines onto orthog-
onal families of confocal hyperbolas and ellipses.
w = cos z = sin
_
z +

2
_
is considered as sine
transformation preceded by translation z +

2
.
Bilinear transformation w =
az+b
cz+d
is conformal ev-
erywhere except at ad bc = 0. It preserves circles.
Cross-ratio is invariant under this transformation. It
has at most two xed points.
The unique bilinear transformation which maps
three given points z
1
, z
2
, z
3
in the z-plane onto three
given points w
1
, w
2
, w
3
in the w-plane is given by
(z
1
z
2
)(z
3
z)
(z
1
z)(z
3
z
2
)
=
(w
1
w
2
)(w
3
w)
(w
1
w)(w
3
w
2
)
.
Schwarz-Christofell transformation determines
functions which conformally maps bounded (or un-
bounded) polygons to upper half plane and conse-
quently to any region into which the half plane can
be transformed such as a unit disk.
Chapter-26
Probability
Probability is a measure of certainty of a probabilis-
tic experiment whose end result or outcome is not
unique.
Sample space is the set of all possible outcomes
of a random experiment. Event is a subset of sample
space
Classical probability P(E) =
Favourable cases
Total cases
has a drawback since the outcomes are assumed to
be equiprobable (equally likely).
Mathematical probability P(E) = lim
n
m
n
may not
have a unique limit.
Axiomatic probability denition is rigorous and
includes the above two denitions also.
i. 0 P(E) 1
ii. P(S) = 1
iii. P(A B) = P(A) +P(B).
Addition theorem for arbitrary events states
P(A B) = P(A) +P(B) P(A B).
Conditional probability of A given B is the proba-
bility of A with respect to the reduced sample space
B and is given by
P(A/B) =
P(A B)
P(B)
.
Multiplication theorem thus states that
P(A B) =P(B)P(A/B) or
=P(A)P(B/A).
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.17
If P(A/B) = P(A) or P(B/A) = P(B) then the
events A, B are said to be independent, in which
case
P(A B) = P(A)P(B).
Theorem of total probability states that for any
event A of sample space S
P(A) =
k

i=1
P(Bi)P(A/Bi)
where B
1
, B
2
, . . . , B
k
form a partition of S.
Bayes Theorem (or rule) determines the proba-
bility of causes
P(Br/A) =
P(Br) P(A/Br)
k

i=1
P(Bi)P(A/Bi)
for r = 1, 2, . . . , k.
Chapter-27
Probability Distributions
A random variable X on a sample space S is a func-
tion X : S R.
Discrete probability distribution of a discrete
RVX is a function f (x) such that f (x) 0 and

f (x) =1
Continuousprobabilitydistributionf (x)satises:
i. f (x) 0
ii.
_

f (x)dx = 1
Chebyshevs theorem on probability is studied.
Discrete uniform distribution dened by
f (x) =
1
k
for X = x
1
, x
2
, . . . , x
k
has mean
1
k
k

i=1
x
i
and variance
1
k
k

i=1
(x
i
).
Binomial distribution is a discrete probability dis-
tribution of the binomial random variable X which
is the number of successes in n Bernoulli trials and
is given by
b(x; n, p) =
_
n
x
_
p
x
q
nx
for x = 0, 1, 2, . . . , n.
The mean and variance are np and npq.
Hypergeometric distribution arises in sampling
without replacement and is dened as
h(x; N, n, k) =
_
k
x
__
N k
n x
_
_
N
n
_ ,
for x = 0, 1, 2, . . . , n.
Its mean is
nk
N
and variance is
nk(Nk)(Nn)
N
2
(N1)
. Hyper-
geometric distribution is approximated by binomial
distribution as N and
k
N
p.
Error function, its properties and its relation to
probability integral are studied.
(Negative) Exponential distribution which has no
memory is useful in reliability theory and queing
theory is dened as
f (x) =
_
e
x
, if x 0
0, ifx < 0
Its mean and standard deviation both are equal to
1

.
Its survival function is
P(x > x) =
_
1 if x < 0
e
x
if x 0
.
The gamma distribution is dened by
f (y) =
_

r
y
r1
e
y
, if y 0
0 , if y < 0
with mean and standard deviation both equal to
r

.
Erlang distribution exponential distribution and
chi-squared distribution are special cases of gamma
distribution.
The weibull distribution is good probability model
for describing length of life of objects having
weakest link property. It is dened as
f (x) =
_

_
xv

_
1
exp
_

_
xv

_
if x v
0 if x < v
Its mean is
_
1

+1
_
+v and variance is

2
_

_
1 +
2

_
1 +
1

__
2
_
2
Summary B.V.Ramana October 13, 2006 12:9
1.18 SUMMARY
Exponential distribution is a special case of both the
gamma and weibull distributions.
Poisson distribution is a discrete probability dis-
tribution given by
f (x, ) =
e

x
x!
, for x = 0, 1, 2, . . .
The mean and variance are both equal to the
parameter . As n and p 0 such that =
np = constant, then binomial distribution can be ap-
proximated by poisson distribution.
Poisson process is a random process in which
the number of events occurring in a time interval
is counted.
Continuous uniform distribution (also known as
rectangular distribution) is given by
f (x) =
_
k = constant, a x b
0, otherwise
Its mean is
b+a
2
and variance is
(ba)
2
12
.
Normal probability distribution also known as
Gaussian distribution is a continuous probability dis-
tribution dened by
N(X, ) =
1

2
e

1
2
(XX)
2
/
2
The graph of normal distribution is known as
the normal curve which is symmetric, bell-shaped,
asymptotic to x-axis on both sides with area under
the curve equals to unity. The probabilities are given
by the area under the normal curve which are tabu-
lated (known as normal tables A.12).
As n and p 0, binomial distribution can
be approximated by normal distribution.
Chapter-28
Sampling Distribution
Population is the totality of observations under study.
Sampleisasubset of population. Statistical quantities
calculated frompopulation are known as parameters,
those calculated from sample as statistics. Statistical
inference deals with the methods of arriving at valid
generalizations and predictions about the population
using the information contained in the sample.
Statistic is a real valued function of the random
sample. It is a random variable and has a probability
or frequency distribution.
Sampling distribution of a statistic is the probabil-
ity distribution of the statistic.
The mean
X
and s.d.
X
of sampling distribution
of means X for:
nite population with sample size n and population
size N:

X
=

X
=

N
_
N n
N 1
,
for innite population (or with replacement)

X
=

X
=

n
Central limit theorem establishes that the sampling
distribution of X (with known) is approximately
normal when n is large (n 30) with standardized
sample mean Z =
X
/

n
.
For small samples (n < 30) S.D. of X is normally
distributed if sampling is from normal population.
S.D. of difference and sums of statistics S
1
and S
2
have
mean =
S
1
S
2
=
S
1

S
2
s.d. =
S
1

S
2
=
_

2
S
1
+
2
S
2
Sampling distribution of X (with unknown) for
small samples follows t -distribution
t =
x
s/

n
with = n 1 dof which approaches normal distri-
bution as n .
Sampling distribution of variance s
2
is related to
the
2
-distribution as follows:

2
=
(n 1)s
2

2
F-distribution determines whether the ratio of two
sample variances from two populations is too small
or too large.
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.19
Chapter-29
Estimation and Test of Hypothesis
Statistical estimation is a part of statistical inference
where a population parameter is estimated from the
corresponding sample statistics. In point estimation
the parameter is estimated by a single number from
sample observations with a maximum error of esti-
mate E = z
/2

n
and the sample size is given by
n = (z
/2
/E)
2
. For small samples E = t
/2
s

n
.
Interval estimate is an interval in which the param-
eter lies with (1 ) 100% condence.
Condence interval for (for large-samples
n30)
x z
/2

n
< < x +z
/2

n
Condence interval for (for small samples
n<30)
x t
/2
s

n
< < x +t
/2
s

n
In Bayesian estimation, prior feelings about the
possible values of are combined with direct sample
evidence which gives the posterior distribution of
, approximately normally distributed with mean

1
=
nx
2
0
+
0

2
n
2
0
+
2
and
s.d.
1
=

2
0
n
2
0
+
2
Bayesian interval for is

1
z
/2

1
< <
1
+z
/2

1
Chapter-30
Curve Fitting, Regression and Correlation
Analysis
Curve tting is the method of nding equation of a
curve that approximates a given set of data.
For a given set of N data points (X
1
, Y
1
),
(X
2
, Y
2
) (X
N
, Y
N
) the best tting curve by the
method of least squares is obtained by minimizing
the sum of the squares of the errors (deviations). The
normal equations for a least squares (L.S.) straight
line are given by

Y
i
=Na
0
+a
1

X
i

X
i
Y
i
=a
0

X
i
+a
1

X
2
i
Nonlinear curves such as exponential curve Y =
AB
X
, geometric curve Y = AX
B
, hyperbola Y =
1
a
0
+a
1
X
can be transformed to linear curve (straight
line). L.S. straight line of Y on Xcan be expressed as
y =
_
xy

x
2
_
x
where x = X X, y = Y Y
Curve tting by sumof exponentials is considered
y = f (x) =
n

i=1
A
i
e
ix
The normal equations, in discrete case, for linear
weighted least squares approximation are given by
a
0

w
i
+a
1

w
i
x
i
=
N

i=0
w
i
y
i
a
0

w
i
x
i
+a
1

w
i
x
2
i
=

w
i
x
i
y
i
In the continuous case, the summations are replaced
by integrals wrt x. For nonlinear weighted LS ap-
proximation of
y = f (x) = a
0
+a
1
x +a
2
x
2
the normal equations in the continuous case are
a
0
_
b
a
wdx +a
1
_
b
a
xwdx +a
2
_
b
a
x
2
wdx =
_
b
a
wy dx
a
0
_
xwdx +a
1
_
x
2
wdx +a
2
_
x
3
wdx =
_
wxy dx
a
0
_
x
2
wdx +a
1
_
x
3
wdx +a
2
_
x
4
wdx =
_
wx
2
y dx
In the discrete case, the integrals are replaced by
summations with i varying from 1 to N.
Simple linear regression of Y on X establishes the
relation between Y and X by the equation
Y = a
0
+a
1
X
which is used for prediction (estimation).
A(1 )100% condence intervals for the pa-
rameters regression coefcients and are given
respectively by
a
0
t
2
s

_
S
xx
+(nx)
2
nS
xx
< <a
0
+t
2
s

_
S
xx
+(nx)
2
nS
xx
Summary B.V.Ramana October 13, 2006 12:9
1.20 SUMMARY
and a
1
t
2
s

s
xx
< < a
1
+t
2
s

s
xx
In curvilinear regression, for a polynomial curve of
degree N
Y = a
0
+a
1
X +a
2
X
2
+ +a
N
X
N
(1)
the (N +1) normal equations to solve for the
(N +1) unknowns a
0
, a
1
, a
2
, . . . , a
N
are obtained
by formally multiplying the regression polynomial
(1) respectively by 1, X
i
, X
2
i
, . . . , X
N
i
and summing
up all the terms from i = 1 to N.
The normal equations in linear multiple regression
for the equation
Y = b
0
+b
1
X +b
2
X
2
are given by

Y
i
=Nb
0
+b
1

X
1i
+b
2

X
2i

X
1i
y
i
=b
0

X
1i
+b
1

X
2
1i
+b
2

X
1i
X
2i

X
2i
Y
i
=b
0

X
2i
+b
1

X
1i
X
2i
+b
2

X
2
2i
Correlation coefcient r measures the strength of re-
lationship between two or more variables. It is given
by the Karl Pearson product-moment formula
r =

xy
_

x
2

y
2
=
N

XY

Y
_
_
N

X
2
(

X)
2
__
N

Y
2
(

Y)
2
_
r is the geometric meanof the regressioncoefcients.
Zero correlation can be tested by the statistic
Z =

n 3
1
2
ln
_
1 +r
1 r
_
Statistic for inference about (= 0) is
Z =

n 3
1
2
ln
_
(1 +r)(1 )
(1 r)(1 +)
_
.
Spearmans rank correlation is a non-parametric
counterpart of the conventional correlation coef-
cient. It is given by
r
rank
= r
S
= 1
6
n

i=1
d
2
i
n(n
2
1)
where d
i
= difference between ranks assigned to
data and n is the size of the data.
For bivariate frequency distribution, the correla-
tion coefcient r takes the form
r =
N

f U
X
U
Y

f
X
U
X
__

f
Y
U
Y
_

_
_
N

f
X
U
2
X

f
X
U
X
_
2
__
N

f
Y
U
2
Y

f
Y
U
Y
_
2
_
(see correlation table on Page 30.34 for notation).
Chapter-31
Joint Probability Distribution and Markov
Chains
In the multivariate case involving two or more ran-
domvariables, the concept of joint probability distri-
bution, marginal distribution and conditional proba-
bility distributions are required. The covariance is
cov(x, y) = E(x, y)
x

y
and correlation is
(x, y) =
cov(x, y)

x

y
A Markov (memoryless) process is a stochastic pro-
cess whose entire past history is summarized in the
present state. The transition matrix P of a Markov
chain is a stochastic matrix. The n-step transition ma-
trix P
n
is equal to P
n
, the nth power of the transition
matrix P.
Chapter-32
Numerical Analysis
The roots of algebra or transcendental equations
are obtained by bisection, regula-falsi and Newton-
Raphson methods.
Finite forward and backward differences and
differences of generalized power are introduced.
Summary B.V.Ramana October 13, 2006 12:9
SUMMARY 1.21
Newton-Gregory forward and backward interpola-
tion formulae are used for interpolation and extrap-
olation. Stirling and Bessels central differences in-
terpolation formulae are used for interpolation near
the middle of the table. Lagranges interpolation is
applied for unequally spaced arguments and is also
used for inverse interpolation. Newtons divided dif-
ferences also can be used for arbitrarily specied
points. The next-term rule is used for estimation of
error in polynomial interpolation. Relationships be-
tween the symbolic operators is considered.
Newtons forward and backward formulae are
used for numerical differentation. The composite,
multi segment or generalized trapezoidal rule, Simp-
sons
1
3
rule, Simpsons
3
8
rule, Weddles rule and
Booles rule are derived and used for numerical in-
tegration.
Cubic polynomials, known as spline functions, are
used in piecewise polynomial approximation. In ad-
dition to cubic splines, linear and quadratic splines
are also considered.
Gauss-Seidel method is an iterative method to get
approximate solution of a system of n equations in
n unknowns.
The largest eigen value and the corresponding
eigen vector of a square matrix is obtained by power
method. Subsequently the smallest and the interme-
diate eigen values are also obtained.
Chapter-33
Numerical Solutions of ODE and PDE
Taylors series, Eulers and modied Eulers method
are single step methods giving numerical solution of
rst order ODE. Runge-Kutta 4th order and Milnes
predictor-corrector methods are multi-step methods
for solution of rst order ODE. Picards method of
successive approximation gives analytical for solu-
tion to rst order ODE. Adams-Bashforth-Moulton
method is another multi step method which is used
more often.
Numerical solution to one-dimensional heat equa-
tion is obtained by Bender-Schmidt recurrence re-
lation. A nite difference approximation is used to
nd the numerical solution if one-dimensional wave
equation. Standard ve point formula, ve point di-
agonal formula are used to obtain numerical solution
of two-dimensional Laplaces solution.

You might also like