Hierarchical Multiple Linear Regression

[Emily Ciesielski]

I am brand new to Stata and trying to finish my dissertation, so I'm hoping someone can help.
I need to run some hierarchical multiple linear regressions, but the data used a multi-stage sampling procedure so I need to account for the lack of a simple random sample-- which I've learned can be done with the command SVY, but "hireg" does not work with SVY commands. All of my variables are continuous.
Thank you!
--Emily

[Scott Merryman]

I think you want -mixed-

Take a look at :
https://www.stata.com/features/overv...s-survey-data/

and

https://www.stata.com/features/overv...h-survey-data/

[Emily Ciesielski]

Thank you for your response! I'm pretty sure I don't need mixed as my data is actually not nested (I'm using older grades so by that time the kids had moved around so much there was no longer a clustering by school). I think I need basically step-wise regression, but it's theory based so it's renamed "hierarchical".
Thanks again!

[Leonardo Guizzetti]

Mixed isn't exclusively for nested, hierarchical regression. It will fit the types of mixed effects models that are commonly called multilevel, hierarchical (I think Raudenbush and Bryk coined this term), crossed effects, and more. That said you will also need to account for the survey design somehow.

[Emily Ciesielski]

Mixed isn't exclusively for nested, hierarchical regression. It will fit the types of mixed effects models that are commonly called multilevel, hierarchical (I think Raudenbush and Bryk coined this term), crossed effects, and more. That said you will also need to account for the survey design somehow.

Thank you! Do you happen to have any resources that I could go to to figure out if a mixed model would work? It has been several years since I've taken my stats classes (I had a few year hiatus in my PhD journey). Basically, I'm trying to determine which method of assessment of executive function "best" predicts success in an academic subject (e.g., reading). My data isn't multi-level (just child-level) but it does come from a multi-stage sampling process.

[Leonardo Guizzetti]

Sorry, I don't usually work with surveys. You should be able to contact a statistician associated with your survey or department to figure out how to account for the design.