Metainformation
| Tag | Value |
|---|---|
| file | Inferential_Statistics_eur-inferential_statistics-103-en_eur-inferential_statistics-103-en |
| name | eur-inferential_statistics-103-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 |
| Program | |
| Language | English |
| Level | Statistical Literacy |
Question
A researcher investigates the criterion validity of a vocabulary test administered in grade 1 in order to predict reading comprehension in grade 4. From previous research she knows that there may be some test bias with respect to the native language of parents. Children whose parents are not Dutch native speakers are expected to show a different relationship between vocabulary and reading comprehension than children whose parents are native Dutch speaker. The results of the regression analyses (n=74, 12% non‐native parents) are given below. Note that the dummy variable “Native Dutch Parents” is 1 when the parents are Dutch native and 0 otherwise.
Explain why the test bias in this case refers to slope bias and not per se to intercept bias.
Solution
The researcher expects that the relationship between vocabulary test and reading comprehension may differ for the two groups. This refers to the slope of the function, and not to the intercept, as the slope is a good indicator of the relation between two quantitative variables. After all, the slope provides information about the following: how much does (predicted) y (in this case reading comprehension in grade 4) changes when x (vocabulary level in grade 1) changes by 1 point?. Also note that the main effect of Native Dutch parents is not significant.