Please I am new to STATA and new to the topic under study for my project as well but happened to have found myself modeling using Mixed Effect Logistic Regression where I built both the Random Intercept and Random Slope Models. The estimates associated with the Random Intercept model comparing the AIC and BIC was identified the best and so I decided to use the Random Intercept model. Now I am confronted with some simulations and the power of these two tests with respect to the two models in fitted where the study specifically focuses on. I will also be glad if I can get the steps that will allow me to resolve this problem in STATA version 11 and possibly if the model fitted is good or otherwise.
The commands use in STATA 11 are as follows
Random Intercept:
noi xtmelogit status i.gender_1 i.age i.typeofschoolatjhs_1 i.regioncompletedatjhs_1, constant||programofstudy_1:, variance nolog
estimates store r_intercept
Random Slope:
noi xtmelogit status i.gender_1 age i.typeofschoolatjhs_1 i.regioncompletedatjhs_1, noconstant||programofstudy_1:gender_1, covariance (unstructured)variance nolog
estimates store r_gender_1
noi esttab r_intercept r_gender_1
estat recovariance
The variable descriptions are as follows
The dependent var is status (1=Pass, 0=Fail)
The predictors are:
gender_1(1=Male, 0=Female)
age_1(0=less than 19years 1= 19-22 years and 2=greater than 22 years)
typeofschoolatjhs_1(0=Private, 1=Public)
regioncompletedatjhs_1(0=Southern, 1=Northern)
The ouptput for the two models are as shown below for N=1940 and 8 grouping variables(programofstudy_1)
The commands use in STATA 11 are as follows
Random Intercept:
noi xtmelogit status i.gender_1 i.age i.typeofschoolatjhs_1 i.regioncompletedatjhs_1, constant||programofstudy_1:, variance nolog
estimates store r_intercept
Random Slope:
noi xtmelogit status i.gender_1 age i.typeofschoolatjhs_1 i.regioncompletedatjhs_1, noconstant||programofstudy_1:gender_1, covariance (unstructured)variance nolog
estimates store r_gender_1
noi esttab r_intercept r_gender_1
estat recovariance
The variable descriptions are as follows
The dependent var is status (1=Pass, 0=Fail)
The predictors are:
gender_1(1=Male, 0=Female)
age_1(0=less than 19years 1= 19-22 years and 2=greater than 22 years)
typeofschoolatjhs_1(0=Private, 1=Public)
regioncompletedatjhs_1(0=Southern, 1=Northern)
The ouptput for the two models are as shown below for N=1940 and 8 grouping variables(programofstudy_1)
| Random-Intercept Model | Random-Slope Model | |||||
| Variables | Coeff | Std Error | Sig. | Coeff | Std Error | Sig. |
| Fixed Effects Estimates | ||||||
| gender_1 | ||||||
| Male | ||||||
| Female | -1.67 | 0.22 | 0.000 | -1.73 | 0.28 | 0.000 |
| age_1 | ||||||
| Below 19 years (*) | ||||||
| 19-22 years | -0.32 | 0.20 | 0.117 | -0.31 | 0.20 | 0.121 |
| Above 22 years schoolstatusatjhs_1 |
-0.42 | 0.31 | 0.182 | -0.42 | 0.31 | 0.175 |
| Public | ||||||
| Private | 0.31 | 0.15 | 0.043 | 0.33 | 0.16 | 0.035 |
| regioncompletedatjhs_1 | ||||||
| Southern | ||||||
| Northern | 1.00 | 0.28 | 0.000 | 1.01 | 0.28 | 0.000 |
| Intercept | -0.27 | 0.39 | 0.490 | -0.33 | 0.44 | 0.450 |
| Random Effects Estimates | ||||||
| Variance (gender_1) | 0.21 | 0.28 | ||||
| Standard Deviation (gender_1) | 0.46 | 0.31 | ||||
| Variance (Constant) | 0.24 | 0.14 | 0.51 | 0.47 | ||
| Standard Deviation (Constant) | 0.49 | 0.15 | 0.71 | 0.33 | ||
| Covariance (gender_1, Constant) | -0.28 | 0.37 | ||||
| Correlation (gender,_1 Constant) | -0.86 | 0.33 | ||||
| Model Fitting Criteria | ||||||
| Log-Likelihood | -1117.76 | -1116.83 | ||||
