STAT802 – Lab exercises – Model answers – Q1
Exercise 1. Consider the data set ‘binary.csv’.
The data set has a binary variable called admit, which is equal to 1 if the individual was admitted
to graduate school, and 0 otherwise. There are three predictor variables: gre, gpa, and rank. We
will treat the variables gre and gpa as continuous. The variable rank takes on the values 1 through
4. Institutions with a rank of 1 have the highest prestige, while those with a rank of 4 have the
lowest. We start out by looking at some descriptive statistics.
Objective 1: To determine to extent to which gpa can be used to predict gre scores, considering
whether the individual was admitted to graduate school. Then, explore the influence of the
variable rank w.r.t gpa only.
Objective 2: To propose a modelling framework to investigate the chances to be admitted to
school upon: gre, gpa, and rank. In other words, to use gre, gpa, and rank to predict admit.
Question 1- Regarding Objective 1:
a) Perform data exploration. Use, e.g., PROC BOXPLOT, PROC FREQ or PROC TABULATE.
Write down 5 – 6 sentences summarizing your results (No Executive summary required).
b) Propose a proper regression model to address this objective. Justify your answer.
c) Write down the theoretical model. Derive the reduced models.
d) Estimate the model using SAS. Write down the estimated model.
e) Write down an executive summary.
ANSWERS:
a) The data set contains 400 observations (prospective students) and four variables: admit, gpa,
gre, and rank. The variables are read in properly by SAS (according to PROC CONTENTS).
Nearly 15% of the observations come from Rank 1 Institutions (highly prestigious), 38% belong
to Rank 2, around 30% are from Rank 3 institutions and 17% are ranked 4. Also, according to the
data, nearly 69% of the individuals were admitted to the graduate school.
COMPLETE THE ANSWER for a) – Do it yourself
b) Regression models for prediction (predictive regression models), as we want to assess the
ability of gpa scores, admission to graduate school, and institution’s rank to predict GPA scores–
An ordinary regression model with normal errors is proposed initially.
A few notes (justifying):
1) We have been asked to “explore the influence of the variable rank w.r.t gpa only”. While
this may be confusing, it simply refers to look at the way the variables ‘rank’ and ‘gpa’
interact in the model. Put another way: Does the GRE difference by rank differ based on
GPA level? This means to explore the effect of GPA over GRE across the different
institution ranks
(variable ‘rank’). For this reason, (solution) the interaction between the
institution’s rank and GPA will be added to the model.
2) We have been asked to “consider whether the individual was admitted to ‘graduate
school’”. This means to check the degree to which the model GRE ~ GPA is altered in the
presence of the variable ‘admit’. Many books refer to this addition to the model as
STAT802 – Week 3 Model Answers – Q1 Due: —
STAT802 – Lab exercises – Model answers – Q1
“exploring the relationship between GPA and GRE adjusted by the variable ‘admit’ “ and
refer to these additional variables (such as ‘admit’) as ‘covariates’.
Solution: The variable ‘admit’ will be added to the linear predictor as COVARIATE.
c) Theoretical full model
GRE | x =
The post STAT802 – Lab exercises appeared first on My Assignment Online.