BUSN 4000 Exam 2
Multicollinearity - CORRECT ANSWER-Any VIF larger than 10 means there is
collinearity present
Large F value, small t values
Prediction vs. Confidence intervals - CORRECT ANSWER-Prediction interval is
ALWAYS wider than confidence interval
This is because average or mean values are always more stable than individual values.
Confidence interval is for the prediction of an average/mean value, whereas prediction
interval is for the prediction of a single value.
Coefficient of Determination - CORRECT ANSWER-R-squared
R^2= SSR/SST
the proportion of variance in Y that is predicted by our model
t-test statistic when testing a specific value - CORRECT ANSWER-b1-B1/sb1
MSR - CORRECT ANSWER-MSR=SSR/k
MSE - CORRECT ANSWER-MSE=SSE/n-k-1
Constructing a confidence interval - CORRECT ANSWER-(b1-t*x sb1, b1+t*x sb1)
where:
b1 = regression coefficient
t* = T.INV((1-% confidence we are testing/2), residual df)
SSE - CORRECT ANSWER-Sum Square due to Error
sum(y-yhat)^2
Outliers - CORRECT ANSWER-Extreme Y value
How to detect: Standard residuals +-2
Leverage Points - CORRECT ANSWER-Extreme X value
How to detect: Highest leverage values from SmartREG
Influential Observations - CORRECT ANSWER-An observation that has a strong
influence on the regression model
How to detect: Highest Cook's D or DFFITS
Multicollinearity - CORRECT ANSWER-Any VIF larger than 10 means there is
collinearity present
Large F value, small t values
Prediction vs. Confidence intervals - CORRECT ANSWER-Prediction interval is
ALWAYS wider than confidence interval
This is because average or mean values are always more stable than individual values.
Confidence interval is for the prediction of an average/mean value, whereas prediction
interval is for the prediction of a single value.
Coefficient of Determination - CORRECT ANSWER-R-squared
R^2= SSR/SST
the proportion of variance in Y that is predicted by our model
t-test statistic when testing a specific value - CORRECT ANSWER-b1-B1/sb1
MSR - CORRECT ANSWER-MSR=SSR/k
MSE - CORRECT ANSWER-MSE=SSE/n-k-1
Constructing a confidence interval - CORRECT ANSWER-(b1-t*x sb1, b1+t*x sb1)
where:
b1 = regression coefficient
t* = T.INV((1-% confidence we are testing/2), residual df)
SSE - CORRECT ANSWER-Sum Square due to Error
sum(y-yhat)^2
Outliers - CORRECT ANSWER-Extreme Y value
How to detect: Standard residuals +-2
Leverage Points - CORRECT ANSWER-Extreme X value
How to detect: Highest leverage values from SmartREG
Influential Observations - CORRECT ANSWER-An observation that has a strong
influence on the regression model
How to detect: Highest Cook's D or DFFITS
