Hello everyone,
As described above, I would like to compare two correlation coefficients from two linear regression models that refer to the same dependent variable (i.e. different x-variables, same y-variable). Is there a command in Stata to test whether the two correlation coefficients (or anything else that could be derived from a linear regression model, such as slope or angle) are significantly different (null hypothesis: the two coefficients are not different)?
Thank you in advance, best regards.
Hello Matteo,
Maybe this can help you:
https://ideas.repec.org/c/boc/bocode/s407302.html
https://www.statalist.org/forums/forum/general-stata-discussion/general/1347283-how-to-compare-two-correlation-coefficients-within-the-same-sample
Such information might be of dubious value, especially if say there is curvature involved, though that could still be considered linear.
Perhaps you should look at "graphical residual analyses," with both models on the same scatterplot for comparison of fit. Cross-validation could be used to consider overfitting. Heteroscedasticity should also be considered in each case. See https://www.researchgate.net/publication/333659087_Tool_for_estimating_coefficient_of_heteroscedasticityxlsx under
https://www.researchgate.net/project/OLS-Regression-Should-Not-Be-a-Default-for-WLS-Regression
for regression of form y = y* + e.
By the way, if you use an hypothesis test, you should never stop with an isolated p-value (rather meaningless), and the threshold should not just be arbitrarily set at a predetermined number (such as 0.05). You need to consider effect size. This is often not very easily interpretable.
- Badges
- Science topic
- Similar topics
- Psychology