I am currently working on my first project using STATA. I used mixed, trying to calculate a multilevel regression with company data (Level 1) and country data (Level 2). In my random slope model the calculation of standard errors failed and the lrtest failed as well stating that the models are not nested. I really tried to find my mistake, but I could not find any help until now.
Short information regarding my model: Turnover_adjusted is the dependent variable (on level 1). Pay based on individual performance (payindivperf) is the independent variable on level 1. The control variables (industryc, sectorc, marketgrc, changec, lnsizec, internationalizationc) are all level 1 variables and grand mean centered. Payindivper_x_appraisal_paybasis is a level 1 moderator. Additionally, we have a cross-level interaction in model 4 with payindivper_x_c_globe_practices. Globe_practices is a level 2 variable. Country_num is also a level 2 variable.
Does anyone has an idea why the calculation of standard errors fails for model 3 and what could be the reasons STATA says that the models 2 and 3 are not nested? For the other models standard error calculation and lrtest were successful. I would be thankful for any help or advice.
Thank you in advance!
Lisa
Code:
mixed turnover_adjusted || country_num:
Code:
estimates store m1
Code:
mixed turnover_adjusted industryc sectorc marketgrc changec lnsizec internationalizationc payindivperf appraisal_paybasis payindivper_x_appraisal_paybasis c_globe_practices || country_num:
Code:
estimates store m2
Code:
mixed turnover_adjusted industryc sectorc marketgrc changec lnsizec internationalizationc payindivperf appraisal_paybasis payindivper_x_appraisal_paybasis || country_num: payindivperf, covariance(unstructured)
Array
Code:
estimates store m3
Code:
lrtest m2 m3
Code:
mixed turnover_adjusted industryc sectorc marketgrc changec lnsizec internationalizationc payindivperf appraisal_paybasis payindivper_x_appraisal_paybasis c_globe_practices payindivper_x_c_globe_practices || country_num: payindivperf, covariance(unstructured)
Code:
estimates store m4
0 Response to Multilevel regression - problem lrtest and standard errors
Post a Comment