## Jointly test coefficient is 0 across 3 outcomes from trivariate cross-classified model fitted using MCMC

Welcome to the forum for runmlwin users. Feel free to post your question about runmlwin here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Go to runmlwin: Running MLwiN from within Stata >> http://www.bristol.ac.uk/cmm/software/runmlwin/
RachaelHughes
Posts: 9
Joined: Thu Jun 13, 2019 10:33 am

### Jointly test coefficient is 0 across 3 outcomes from trivariate cross-classified model fitted using MCMC

Hi,

I've fitted a trivariate cross-classified model using MCMC.

I've been asked to calculate classical p-values for the joint test that a fixed-effect coefficient in the model equals 0 for all 3 dependent variables. For example, after fitting a nested model using RIGLS, I can use STATA command test to jointly test if the coefficient for covariate dr is 0 across the three dependent variables:

test dr_1 dr_2 dr_3

( 1) [FP1]dr_1 = 0
( 2) [FP2]dr_2 = 0
( 3) [FP3]dr_3 = 0

chi2( 3) = 310.87
Prob > chi2 = 0.0000

Is it okay to calculate this by hand for the trivariate cross-classified model (estimated using MCMC)? So using the relevant fixed effect estimates b and the relevant part of the fixed effect covariance matrix V I can calculate the chi-squared statistic as b*invsym(V)*b' and the get the 2-sided p-value from the the reverse cumulative chi-squared distribution with 3 degrees of freedom (e.g., 2*chi2tail(teststatistic)).

Is this a reasonable thing to do for a trivariate cross-classified model estimated using MCMC?

Thanks
Rach
billb
Posts: 140
Joined: Fri May 21, 2010 1:21 pm

### Re: Jointly test coefficient is 0 across 3 outcomes from trivariate cross-classified model fitted using MCMC

Hi Rachael,
MCMC is fitting a Bayesian model so the concept of a classical P value is not valid for MCMC estimates - I think we do offer Bayesian P values for all parameters though which basically check what proportion of iterations are positive / negative for a chain.
Best wishes,
Bill.