Has anyone used R software to perform a hierarchical linear regression model? There are three levels in my research: school, class, and children. I made a preliminary attempt to determine whether three levels need to be included in the model. However, when I tried from the three levels-lmer<-formula~1+(1|school)+(1|class)+(1|children), I found that the results in the model always reported errors:
Error in eval_f(x, …) : Downdated VtV is not positive definite.
When I removed the smallest level, it became a two-level model-lmer<-formula~1+(1|school)+(1|class). The warning still appears:
Warning messages:
1: In checkConv(attr(opt, “derivs”), opt
p
a
r
,
c
t
r
l
=
c
o
n
t
r
o
l
par, ctrl = control
par,ctrl=controlcheckConv, :
unable to evaluate scaled gradient
2: In checkConv(attr(opt, “derivs”), opt
p
a
r
,
c
t
r
l
=
c
o
n
t
r
o
l
par, ctrl = control
par,ctrl=controlcheckConv, :
Model failed to converge: degenerate Hessian with 1 negative eigenvalues
3: Model failed to converge with 1 negative eigenvalue: -1.0e-01
When I continued to reduce the number of levels and nearly included the random term school (1|school)in the model, the model no longer displayed errors and warnings.
I don’t know what went wrong, I have no way to verify the model of three random terms. Please enlighten me.
Multilevel linear regression modelR语言多级线性回归模型
最新推荐文章于 2024-05-10 17:01:28 发布