BUAN 6392 Midterm Questions with Correct answers
Point Estimate of Population ATE Correct Answer-Difference in means
of outcome of control and treatment group
Fairness of ATE estimate Correct Answer-No selection bias, exclusion
restriction, non-interference
As small a sampling variance as possible
Small sampling variance = better likelihood to find treatment effect
Exclusion restriction Correct Answer-The reason for the difference
between behavior of treatment and control groups is the treatment and
nothing else
Non-interference Correct Answer-The value of the potential outcome for
a unit depends only on whether it received the treatment, and not on
treatment received by other units. Also called SUTVA (Stable Unit
Treatment Value Assumption)
Hypothesis Testing using t -statistic Correct Answer-• Total number of
observation in treatment + control groups = N • Number of observation
in treatment group = m
• Number of observations in control group = N - m
Estimated SE of ATE Correct Answer-
T statistic Correct Answer-
, Design for high precision of ATE Estimator Correct Answer-•
Increasing size of control group (for fixed m) increases N, reducing SE
• Increasing size of treatment group m (for fixed N-m), reduces SE
• So, increase both control and treatment group when possible (less
costly)
Confidence Interval Correct Answer-•When the goal is to form an idea
about the value of ATE (and not to check that it is different from 0), then
we use confidence intervals
• A 95% CI means that the procedure used to generate the CI will
contain the true ATE for 95% of the samples
R code for t-test Correct Answer-
t.test(df$sales[df$website=="B"],df$sales[df$website=="A"])
Linear Regression Model Correct Answer-
R Code: Regression website design on sales Correct Answer-
summary(lm(formula = sales ~ website, data = df))
R code: Regressing website design on log sales Correct Answer-
summary(lm(formula = log(sales)~website, data = df))
Point Estimate of Population ATE Correct Answer-Difference in means
of outcome of control and treatment group
Fairness of ATE estimate Correct Answer-No selection bias, exclusion
restriction, non-interference
As small a sampling variance as possible
Small sampling variance = better likelihood to find treatment effect
Exclusion restriction Correct Answer-The reason for the difference
between behavior of treatment and control groups is the treatment and
nothing else
Non-interference Correct Answer-The value of the potential outcome for
a unit depends only on whether it received the treatment, and not on
treatment received by other units. Also called SUTVA (Stable Unit
Treatment Value Assumption)
Hypothesis Testing using t -statistic Correct Answer-• Total number of
observation in treatment + control groups = N • Number of observation
in treatment group = m
• Number of observations in control group = N - m
Estimated SE of ATE Correct Answer-
T statistic Correct Answer-
, Design for high precision of ATE Estimator Correct Answer-•
Increasing size of control group (for fixed m) increases N, reducing SE
• Increasing size of treatment group m (for fixed N-m), reduces SE
• So, increase both control and treatment group when possible (less
costly)
Confidence Interval Correct Answer-•When the goal is to form an idea
about the value of ATE (and not to check that it is different from 0), then
we use confidence intervals
• A 95% CI means that the procedure used to generate the CI will
contain the true ATE for 95% of the samples
R code for t-test Correct Answer-
t.test(df$sales[df$website=="B"],df$sales[df$website=="A"])
Linear Regression Model Correct Answer-
R Code: Regression website design on sales Correct Answer-
summary(lm(formula = sales ~ website, data = df))
R code: Regressing website design on log sales Correct Answer-
summary(lm(formula = log(sales)~website, data = df))
