A Stata MLM Workflow

stats
teaching
workflow
Stata
multilevel
Author

Andy Grogan-Kaylor

Published

July 1, 2025

Background

I’m sometimes asked for a quick outline of how to proceed with a multilevel analysis. Below are the steps that I would use in Stata, though they would be very conceptually similar in R.

Get acquainted with your data

Figure out what variables correspond to what measures and constructs. e.g.:

  • lookfor health
  • lookfor ses
  • describe question37

Prepare and clean your data

Rename variables to be more intuitive. e.g.:

rename q21 health

Recode missing values. e.g:

recode health (99=.)

Keep only necessary variables.

keep age race sex ses health q21 neighborhood

Descriptive statistics

  • summarize continuousvars
  • tab1 categoricalvars

And in more recent versions of Stata

  • dtable continousvars i.categoricalvar1 i.categoricalvar2

Multilevel model

  1. mixed y || groupid: // unconditional model
  2. estat icc
  3. mixed y x1 x2 || groupid:
  4. mixed y x1 x2 || groupid: x1 x2

Graph

sample Stata graph

sample Stata graph

  • histogram y

  • twoway scatter y x1

  • spagplot y x1 id(groupid)