1
$\begingroup$

Here are three regression equations:

$$ \begin{align} Y &= \beta_1Z_{\text{math}}\qquad (1)\\ Y &= \beta_2Z_{\text{math}} + \beta_3Z_{\text{science}}\qquad (2)\\ Y &= \beta_4Z_{\text{department}} + \beta_5Z_{\text{math}} + \beta_6Z_{\text{science}} + \beta_7Z_{\text{university}} + \beta_8Z_{\text{gender}}\qquad (3) \end{align} $$

I have 3-level regression model. I want to control the effect of maths achievement alone for (maths and science achievement in model 2), and model 3 includes gender, university and department.

B1math is SS in Model 1, B2science is SS in model 2, and B7university and B6science is SS in model three. I have 15, 25, and 35 R-squared estimates.

I am not sure what to report, besides increase in R-squared? Should I report the standardized Beta weight of university alone (because science was controlled) or should I also report B of science?

Do structure coefficient play a role in interpreting B?

$\endgroup$
1
$\begingroup$

If you are writing for a journal, they will probably have a style guide. I, personally, would make a table with each model and the associated $\beta$s, their SEs, and CIs. Most places will require p-values (but by choice I would leave them out). Standardized betas are required by the Am. Psychological Association style, but I don't like them.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks Peter. But, in the final model, should I discuss the beta weight of science, which is SS but was controlled in previous models? If I do have to discuss, what is the point of doing a hierarchical regression but not simple multiple regression with all bundled in one equation? $\endgroup$ – JonBonJovi Feb 1 '12 at 21:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.