I have done hierarchical regression analysis. My model 1 and model 2 both are significant. But the beta values for predictors in Model 1 are not significant. How to report this?
The most common format is to generate a table where the columns are the two models and the rows are independent variables. Your order the independent variables so that the first set of rows contain the variables that are present in both models, followed by the variables that are present only in the second model.
For your first model, if each variable makes a small but nonsignificant contribution, then you could have a total explained variance that is significant. In that case, I would expect to see a low R-sq, which gets substantially higher when you add the variables in the second model.
I am not sure I completely understand the situation here. Are you saying that in your level one model (fixed effects), none of the covariates are statistically significant (
Just to be clear, my advice was for a "hierarchical comparison" between two regression models, rather than for a Hierarchal Linear Model (Multi-Level Modeling).
Good point, David. I think this alludes to the need for more details on what actually was done.
Dear Dr David and Ariel,
I have attached the table. You will find that my model 1 is significant however beta values for predictors are not. So I want to know how to report this ?
I agree with Respected David Morgan Sir. Since R2 Change and F is significant but beta value are not significant. It can be because R2 Change is significant at .044 and age group, education and gender were explaining only 3.9 % of total variance in social. You can say that these variables were not found to be significantly negatively associated with social.
To avoid confusion, you might refer to what you are doing as "a series of hierarchically ordered regression equations."
Dear Dr David ,
Could you please give explanation for this ? That will be very helpful for me. Thanks in advance.
As I noted earlier, there is a difference between hierarchically ordered regression equations and the techniques that go by the name Hierarchal Linear Models (Multi-Level Modeling). In particular, the advice that Ariel gave you was for the second of these two options, so saying "hierarchically ordered regression equations" helps avoid that potential confusion.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Values