multi-level modeling with a fractional response variable

[Elena Se]

Dear members,

I would like to ask whether it is possible to estimate a multi-level model with a fractional response variable in Stata. In my data, the variable of interest is a percentage (between 0 and 1) nested in countries. As far as I understand, Stata allows to model fractional response regression with clustered standard errors by country using the fracreg command. Among multi-level models, I found meglm that is possible to use with a fractional variable, but binomial family cannot handle 1s in the data. So, I need to transform the dependent variable in order to be able to use it. Are there any other options to run a multi-level model with a fractional response variable?

Thank you for your suggestions!

[Andrew Musau]

Have a look at Papke and Wooldridge (2008). You could use xtgee with a probit link function and an unstructured within-group correlation structure to implement their suggested estimation procedure.


Reference:
Papke, L.E. and J.M. Wooldridge (2008). Panel Data Methods for Fractional Response Variables with an Application to Test Pass Rates," Journal of Econometrics 145, 121-133. https://www.sciencedirect.com/scienc...0440760800050X