Master of Science II (Statistics) Examination: Oct Nov 2016
Semester IV (New CGPPA)
SLR No. Day
Date Time Subject Name Paper
No. Seat No.
SLR SB
705
Thursday
17/11/2016
02.30 PM
to
05.00 PM
Discrete Data Analysis
C
XVI
Instructions: Answer any five questions.
Q. No. and Q. No. are compulsory.
Attempt any three from Q. No. to Q. No.
Figures to the right indicate full marks.
Total Marks:70
Q.1 A. Select the correct alternative: 05
In a 2x2 log linear model, if one observed cell frequencies is zero, then
the maximum likelihood estimate of this cell.
is always zero may or may not zero
is never zero may be negative
Which of the following is not a particular care of generalized linear
models
Poisson regression Classical linear regression
Logistic regression None of these
Logistic regression model is an appropriate when the response variable
as
Poisson Gamma
Normal Binomial
The degrees of freedom associated with the U12 term in log-linear model
for an I x J table are
IJ IJ 1
I J IJ J 1
Which of the following measures is used to test the goodness of fit
of GLM?
t-statistic F-ratio
Deviance Name of these
B. Fill in the blanks: 05
The cross product ratio in log-linear model is also known as


The transformation is called transformation.
In GLM the response variable is assumed to be a member of

statistic is used for testing goodness of fit in a GLM.
In regression analysis when outcome variable is dichotomous, E
lies in to
Page 1 of 2
C. State whether following statements are true or false: 04
A logistic regression is suitable to investigate the relationship between
average milage and weight of the car.
The cross product ratio is always invariant under row and column
transformation.
A saturated model for I X J X k table contains less than I X J X K
independent parameters.
Poisson regression is applicable when response variable dichotomous.
Q.2 A. Explain the terms: 06
Poisson sampling scheme
Generalized linear model
B. Write short note on the following 08
Hierarchical family of models
Poisson regression
Q.3 A. Establish relationship between the cross product ratio and the U12 term in a
log-linear model for a 2 x J table.
07
B. Consider a log-linear model for I X J X K table with U123 0 obtian set of
minimal sufficient configuration.
07
Q.4 A. Derive the maximum likelihood estimates of parameters in GLM. 07
B. Discuss the link function and deviance is the context of GLM. 07
Q.5 A. What is logistic regression model? Give two situations where such a model
is appropriate?
07
B. Derive the maximum likelihood estimator of the parameter of a logistic
regression model with one covariate.
07
Q.6 A. Starting from a log linear model for a two dimensional table explain how a
model for three dimensional table can be developed.
07
B. Define one parameter exponential family obtains its mean and variance. 07
Q.7 A. Define log-odds ratio. Describe its computational procedure with reference
to logistic regression.
07
B. Describe the iterative proportional fitting algorithm for fitting log-linear
model.
07
Page 2 of 2