Skip to content

Book Info
- About
- Preface
- Contents
- Buy
- Companion Sites
- Corrections
- Author
Learning Resources
Teaching Resources
- Presentation Slides
- Instructor’s Manual
Software Info
- Data Desk
- EViews
- JMP
- Minitab
- R
- SAS Code
- SAS Studio
- SPSS
- Stata
- Statistica
Download Data

Book Info
- About
- Preface
- Contents
- Buy
- Companion Sites
- Corrections
- Author
Learning Resources
Teaching Resources
- Presentation Slides
- Instructor’s Manual
Software Info
- Data Desk
- EViews
- JMP
- Minitab
- R
- SAS Code
- SAS Studio
- SPSS
- Stata
- Statistica
Download Data

© 2019

Applied Regression Modeling

3rd edition

Iain Pardoe

A concise, comprehensive treatment of the application of statistical regression analysis suitable for undergraduate and graduate students. Covers simple linear regression, multiple linear regression, model building, and advanced regression topics.

on July 30, 2021

Answer the following questions based on Sections 5.2.6-8: Regression pitfalls - extrapolation, missing data, power and sample size.

1.

What is “extrapolation” in the context of regression analysis?

Misinterpreting a prediction interval as a confidence interval or vice versa.

Using a regression model to estimate or predict outside the scope of the model.

Fitting a model with no intercept term.

The model includes unimportant or irrelevant predictors that make the model overly complex.

None

2.

Suppose we’ve fit a simple linear regression model to sample data that has values of the predictor variable, X, between 1 and 10. Consider using the estimated regression equation to make a prediction of Y for X = 20?

This is reasonable because the estimated regression equation is a straight line that is easy to extend beyond the sample data.

This is reasonable if we report a prediction interval, which will reflect increased uncertainty away from the sample mean of X.

This is unwise since we have no data near X = 20 to support a model fit to data with values of X between 1 and 10.

This is unwise if the value of R-squared is below 90%, but fine otherwise.

None

3.

Select the most appropriate description of missing data in the context of regression analysis.

The model is estimated with a portion of the data missing and then validated using the missing portion.

Observations with missing values for some cases can reduce the total usable sample size for the model.

Observations with the response variable missing can be estimated by using the fitted regression equation.

Excluding important predictors can lead to meaningless models with misleading associations.

None

4.

Which of the following are reasonable strategies for mitigating problems caused by missing data? (Select all that apply.)

Double-check the data source to see if missing values have been omitted by mistake.

Replace missing values with random data.

Attempt to build regression models amongst the predictors to impute missing values using predictions based on these regression models for the predictors.

Consider excluding less useful predictors with large numbers of missing values.

5.

How do power and sample size impact regression analysis?

Models lack power if the sample size is too small to be able to detect significant associations between the response and predictors.

Power can be increased by conducting hypothesis tests using a significance level of 0.01 instead of 0.05.

Power and sample size can both be increased by adding more predictor variables to the model.

Power is reduced when estimated model parameters are highly correlated with one another.

None

Time's up

Categories:

No Category

Tags:

No Tag

Post navigation

Previous post Q5.2.5: Regression pitfalls – overfitting

Post navigation

Next post Q5.4: Model selection

Comments are closed

© 2026 Applied Regression Modeling, 3rd edition. Created for free using WordPress and Colibri