M.Sc. (Semester IV) (CBCS) Examination Oct/Nov-2017
Statistics
DISCRETE DATA ANALYSIS
Day Date: Friday, 17-11-2017 Max. Marks: 70
Time: 02.30 PM to 05.00 PM
Instructions: Attempt five questions.
Question No. 1 and 2 are compulsory
Attempt any 3 questions from Q.No.3 to Q.No.7
Figures to the right indicate full marks.
Q.1 Choose the correct alternatives 05
A GLM is defined when response variable belongs to family of
distribution.
Exponential Pareto
Normal Other than these
The complementary log link is given by
Log − log -log − log
log log 1 − log − log 1 −
For I x J table, degrees of freedom for term area)
IJ I J 1 IJ 1
None of these
Overdispersion arise mainly due to
Lack of fit to the model Heterogeneity in the data
Variability in response None of the these
If then we must have
0 0
0 None of these
Fill in the blanks. 05
In a logistic regression model the response variable is
Quasi-likelihood estimation method is used when there is
Wald statistics for testing significance of individual significance follows
distribution.
Saturated model contains parameters.
The logistic regression of Y on X is
State whether following statements are true or false: 04
Exponential distribution with scale and location parameters belongs to
exponential family.
If we interchange rows and columns then odds ratio remains
unchanged.
Logistic regression used for categorical response.
GLM is only used for non-normal response.
Page 2 of 2
SLR-MS-669
Q.2 Define one parameter exponential family. Show that negative
binomial. Distribution is member of this family.
06
What is baseline category model? Write its applications.
Write short notes on the following: 08
Analysis of Deviance
Poisson sampling scheme
Q.3 Define Generalized Linear Model. Obtain ML estimates of parameters of
a GLM.
07
Show that belongs to exponential family, hence obtain its mean
and Variance.
07
Q.4 Write effects of rearranging data on cross product ratio. Hence establish
relationship between cross product ratio and linear constraint for a 2 x J
table.
07
Define Poisson and Multinomial sampling scheme. Hence obtain kernel
function in both cases.
07
Q.5 What is Overdispersion? Discuss causes of overdispersion. 07
Consider log-linear model for 2x2 incomplete tables and obtain estimates
of U-terms of this model.
07
Q.6 Define a log-linear model for 2x2x2 table and interpret the terms in the
model.
07
Write short notes on: 07
1. Saturated model
2. Comprehensive model
3. Hierarchical Principle
4. Relative risks
Q.7 Discuss measures of Goodness of fit of GLM. 07
What is Quasi-likelihood estimation? Obtain expression QL estimates. 07