Regression

Question 0

RSS + SSE = ?

What do they stand for?

Answer 0

= Total variation (SST)

RSS: Regression sum of squares
SSE: Sum of squared errors

Question 1

What does a low p-value indicate about the significance?

Answer 1

Significance is high

Question 2

What does mean regression sum of squares (MSR) and mean squared error (MSE) equal?

Answer 2

MSR = RSS/k
MSE = SSE/(n-k-1)

Question 3

What is the equation for F-Stat?

Answer 3

F = MSR/MSE

Question 4

What is the equation for R squared?

Answer 4

R squared = RSS/SST = SST-SSE/SST

= Explained regression / Total variation

Question 5

Is Ra squared less than or equal to R squared?

Answer 5

Adjusted R squared is less then R squared

Question 6

What phenomenon occurs When the variance of the residuals are not the same across all observations?

Answer 6

Heteroskedasticity. Unconditional occurs when variance is not related the value of the independent variable. Conditional is when it is related to the independent variable.

Question 7

How do you test for heteroskedasticity?

Answer 7

Breusch-Pagan test.

Question 8

What is the formula for the Breusch- Pagan chi-square test?

Answer 8

BP chi-square test = n x Rsquared of residuals.

This is the Rsquared from a second regression of the squared residuals from the first regression on the independent variables.

Question 9

How do you correct for heteroskedasticity?

Answer 9

Either use White-corrected standard errors or use generalised least squares.

Question 10

How do you test for serial correlation?

Answer 10

Durbin Watson statistic. If DW < d1 error terms are positively serially correlated. If d1< DW < du the test in inconclusive. If DW > du there is no evidence that error terms are positively correlated.

Question 11

How do you correct for serial correlation?

Answer 11

Use Hansen method to adjust the coefficient standard errors.

Or improve specification of the model.

Question 12

What is multicollinearity?

Answer 12

When two or more independent variables in a multiple regression are highly correlated with each other

Question 13

When the t-test indicates that none of the individual coefficients are significantly different from zero, F-test is statistically significant and the Rsquared is high what regression problem does this potentially indicate?

Answer 13

Multicollinearity

Question 14

What distribution is a probit model based on?
And what distribution is a logit model based on?
What models are these used for?

Answer 14

Probit - normal distribution
Logit - logistic distribution

Used for qualitative dependent variable models. Results in estimates of the probability that the event occurs.

Question 15

A qualitative dependent variable model that results in a linear function which generates an overall score or ranking for an observation is called what?

Answer 15

Discriminant model

Question 16

What condition exists if the variance of residuals in one period is dependent on the variance of the residuals in a previous period?

Answer 16

Autoregressive conditional heteroskedasticity (ARCH)

Question 17

If both heteroskedasticity and serial correlation are present what method do you use to correct this?

Answer 17

Hansen method which adjusts the coefficient standard errors

Question 18

How do you correct for multicollinearity?

Answer 18

Omit one or more of the correlated independent variables.

Question 19

Is a random walk covariance stationary?

Answer 19

No. Neither a random walk nor a random walk with a drift exhibits covariance stationary.