Random slope About eopletamuedu随机斜率约peopletamuedu.pptVIP

Random slope Varying slopes Basic Model: Varying slopes Gelman and Hill Presentation Varying slopes The really tricky part is that correlation term—higher group level errors in the DV may also be connected to higher (or lower) effects of X. In R it is pretty easy Just add which individual level effect you want to vary as a function of the group Two examples Radon NES Add the second random term by putting the variable before the |”group” Add second level predictors in the random slope by creating an interaction R output Linear mixed-effects model fit by REML Formula: y ~ x + u.full + x:u.full + (1 + x | county) AIC BIC logLik MLdeviance REMLdeviance 2141 2174 -1063 2114 2127 Random effects: Groups Name Variance Std.Dev. Corr county (Intercept) 0.0155 0.124 x 0.0942 0.307 0.409 Residual 0.5617 0.749 number of obs: 919, groups: county, 85 Random effects who the varainces and the correlation Output Fixed effects: Estimate Std. Error t value (Intercept) 1.4686 0.0352 41.7 x -0.6710 0.0844 -7.9 u.full 0.8081 0.0906 8.9 x:u.full -0.4195 0.2271 -1.8 Correlation of Fixed Effects: (Intr) x u.full x -0.241 u.full 0.208 -0.093 x:u.full -0.093 0.173 -0.231 The fixed effects coefficients are the gamma terms Output Output NES output Linear mixed-effects model fit by REML Formula: y ~ x + u.full + x:u.full + (1 + x | district) AIC BIC logLik MLdeviance REMLdeviance 11720 11756 -5853 11722 11706 Random effects: Groups Name Variance Std.Dev. Corr district (Intercept) 92.6 9.62 x 201.1 14.18 -0.542 Residual 951.4 30.85 number of obs: 1199, groups: district, 139 NES Output Fixed effects: Estimate Std. Error t value (