Metainformation
| Tag | Value |
|---|---|
| file | Inferential_Statistics_eur-inferential_statistics-102-en_eur-inferential_statistics-102-en |
| name | eur-inferential_statistics-102-en |
| section | Inferential Statistics/Regression/Equation, Inferential Statistics/Regression/Intercept, Inferential Statistics/Regression/Slope, Inferential Statistics/Regression/Simple linear regression |
| type | string |
| solution | "" |
| Type | Conceptual / Interpreting output |
| Program | |
| Language | English |
| Level | Statistical Literacy |
Question
A researcher wants to investigate the criterion validity of a test anxiety test used to predict the level of experienced stress during an exam. He wonders whether there is an intercept bias due to differential effect of gender on the relationship. The table below shows the SPSS output of the regression analysis the researcher performed (n=52, 49% woman). Note that men were coded 0 and women were coded 1.
Based on the output, what do you conclude about the intercept bias?
Solution
When test anxiety is 0, men’s stress score is 3.601 and women’s stress score is 5.854. The difference at anxiety=0 between men and women is significant (p<.001) so the intercept bias is probably present. Women will tend to have higher stress scores at any anxiety score than men with the same anxiety score. Thus, the predictor anxiety level does not “work” in the same way for the two genders.