\Exam SRM/
The variance of a statistical learning method decreases as the method's flexibility increases - correct
answer ✔✔False. Flexibility increases, Variance increases.
SST - correct answer ✔✔Total Sum of Squares. Sum of (yi - average y)^2
SSR - correct answer ✔✔Regression Sum of Squares. Sum of (y hat - yi)^2
SSE aka RSS - correct answer ✔✔Error Sum of Squares/Residual Sum of Squares. Sum of (yi - y hat)^2
s^2 aka MSE - correct answer ✔✔SSE/(n-p-1)
Residual Standard Error - correct answer ✔✔Square rut of the MSE
R^2 - correct answer ✔✔coefficient of determination. SSR/SST. 1 - SSE/SST
t stat - correct answer ✔✔(b1 - B1)/ seb1
Which interval is always at least as wide as the other? - correct answer ✔✔A Prediction interval is always
at least as wide as the Confidence Interval.
If p-value is less than or equal to alpha - correct answer ✔✔Reject the Null Hypothesis (the variable is
significant)
R^2 adj - correct answer ✔✔1- MSE/variance. 1 - (SSE/(n-p-1))/(SST=(n-1)). 1 - (1-R^2)*((n-1)/(n-p-1))
The sum of residuals always = ________ - correct answer ✔✔p + 1
, A residual (hi) is always between ______ - correct answer ✔✔1/n and 1. (1/n < hi < 1)
FREES: outlier if __________. - correct answer ✔✔standardized residual is > 2
FREES: high leverage point if hi > ___________ - correct answer ✔✔3((p+1)/n)
FREES: Severe multicollinearity if ___________. - correct answer ✔✔VIF greater than or equal to 10.
The logit and probit functions are substantially different. - correct answer ✔✔False.
F-Stat (full) - correct answer ✔✔MSR/MSE. (SSR/p)/(SSE/(n-p-1))
F-stat (Partial) - correct answer ✔✔(SSEr - SSEf)/(pf-pr) / SSEf/(n-pf - 1)
What is the difference between a t test and an F test? - correct answer ✔✔T test: one parameter at a
time. F Test: more than one parameter at a time.
The sum of leverages in a model (hi) = ? - correct answer ✔✔p + 1
How do you find the leverages in a matrix? - correct answer ✔✔They are the diagonal elements of the
Hat matrix
Cook's Distance formula? What is it for? - correct answer ✔✔Measuring leverage and residuals in one
formula. ((ei^2 * hi)/(1 - hi)^2) / MSE(p+1)
Heteroscedasticity - correct answer ✔✔Non-constant variance. Often takes on a funnel shape when
looking at a plot of residuals
What are AIC, BIC, and Mallow's Cp used for? - correct answer ✔✔They summarize the tradeoff between
model fit and complexity. Smallest value is desired.
The variance of a statistical learning method decreases as the method's flexibility increases - correct
answer ✔✔False. Flexibility increases, Variance increases.
SST - correct answer ✔✔Total Sum of Squares. Sum of (yi - average y)^2
SSR - correct answer ✔✔Regression Sum of Squares. Sum of (y hat - yi)^2
SSE aka RSS - correct answer ✔✔Error Sum of Squares/Residual Sum of Squares. Sum of (yi - y hat)^2
s^2 aka MSE - correct answer ✔✔SSE/(n-p-1)
Residual Standard Error - correct answer ✔✔Square rut of the MSE
R^2 - correct answer ✔✔coefficient of determination. SSR/SST. 1 - SSE/SST
t stat - correct answer ✔✔(b1 - B1)/ seb1
Which interval is always at least as wide as the other? - correct answer ✔✔A Prediction interval is always
at least as wide as the Confidence Interval.
If p-value is less than or equal to alpha - correct answer ✔✔Reject the Null Hypothesis (the variable is
significant)
R^2 adj - correct answer ✔✔1- MSE/variance. 1 - (SSE/(n-p-1))/(SST=(n-1)). 1 - (1-R^2)*((n-1)/(n-p-1))
The sum of residuals always = ________ - correct answer ✔✔p + 1
, A residual (hi) is always between ______ - correct answer ✔✔1/n and 1. (1/n < hi < 1)
FREES: outlier if __________. - correct answer ✔✔standardized residual is > 2
FREES: high leverage point if hi > ___________ - correct answer ✔✔3((p+1)/n)
FREES: Severe multicollinearity if ___________. - correct answer ✔✔VIF greater than or equal to 10.
The logit and probit functions are substantially different. - correct answer ✔✔False.
F-Stat (full) - correct answer ✔✔MSR/MSE. (SSR/p)/(SSE/(n-p-1))
F-stat (Partial) - correct answer ✔✔(SSEr - SSEf)/(pf-pr) / SSEf/(n-pf - 1)
What is the difference between a t test and an F test? - correct answer ✔✔T test: one parameter at a
time. F Test: more than one parameter at a time.
The sum of leverages in a model (hi) = ? - correct answer ✔✔p + 1
How do you find the leverages in a matrix? - correct answer ✔✔They are the diagonal elements of the
Hat matrix
Cook's Distance formula? What is it for? - correct answer ✔✔Measuring leverage and residuals in one
formula. ((ei^2 * hi)/(1 - hi)^2) / MSE(p+1)
Heteroscedasticity - correct answer ✔✔Non-constant variance. Often takes on a funnel shape when
looking at a plot of residuals
What are AIC, BIC, and Mallow's Cp used for? - correct answer ✔✔They summarize the tradeoff between
model fit and complexity. Smallest value is desired.
