我是R新手，并尝试使用glmer来帮助完成我的论文。我使用的是Stat v16，但meglm不允许我使用伽马分布和链接标识，因此我来到R来执行此操作。不幸的是，我仍然无法使完整的模型收敛。
不确定这是否有帮助，但这就是我在Stata中试图做的事情：
meglm countOutcome Exposure##c.TimePeriod1 Exposure##c.TimePeriod2 Exposure##c.TimePeriod3 Exposure##c.TimePeriod4 covariate1 covariate2 covariate3, || SBA:, family(gaussian) link(identity) covariance(unstructured) offset(area) reml
这是我尝试过的...
第一次尝试：
summary(glmer(countOutcome ~ 1 + Exposure.f*TimePeriod1 + Exposure.f*TimePeriod2 + Exposure.f*TimePeriod3 + Exposure.f*TimePeriod4 + covariate1 + covariate2 + covariate3 + (1 | SBA), data=data,family=Gamma(link=identity)))
fit warnings:
Some predictor variables are on very different scales: consider rescaling
convergence code: 0
unable to evaluate scaled gradient
Model failed to converge: degenerate Hessian with 4 negative eigenvalues
failure to converge in 10000 evaluations
Warning messages:
1: Some predictor variables are on very different scales: consider rescaling
2: In (function (fn, par, lower = rep.int(-Inf, n), upper = rep.int(Inf, :
failure to converge in 10000 evaluations
3: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
unable to evaluate scaled gradient
4: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge: degenerate Hessian with 4 negative eigenvalues
5: In vcov.merMod(object, use.hessian = use.hessian) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
6: In vcov.merMod(object, correlation = correlation, sigm = sig) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
我通过以均值为中心对协变量进行了缩放，但仍然收到警告。
fit warnings:
Some predictor variables are on very different scales: consider rescaling
convergence code: 0
unable to evaluate scaled gradient
Model failed to converge: degenerate Hessian with 4 negative eigenvalues
failure to converge in 10000 evaluations
Warning messages:
1: Some predictor variables are on very different scales: consider rescaling
2: In (function (fn, par, lower = rep.int(-Inf, n), upper = rep.int(Inf, :
failure to converge in 10000 evaluations
3: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
unable to evaluate scaled gradient
4: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge: degenerate Hessian with 4 negative eigenvalues
5: In vcov.merMod(object, use.hessian = use.hessian) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
6: In vcov.merMod(object, correlation = correlation, sigm = sig) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
然后，我按每1000个人重新调整了协变量的比例，但仍然收到警告。
convergence code: 0
unable to evaluate scaled gradient
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
Warning messages:
1: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
unable to evaluate scaled gradient
2: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
3: In vcov.merMod(object, use.hessian = use.hessian) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
4: In vcov.merMod(object, correlation = correlation, sigm = sig) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
在Stack Exchange上浏览了几篇非常有用的文章之后，我尝试了另外三种方法但无济于事...非常欢迎提出建议。
尝试使用bobyqa作为优化程序：
summary(glmer(countOutcome ~ 1 + Exposure.f*TimePeriod1 + Exposure.f*TimePeriod2 + Exposure.f*TimePeriod3 + Exposure.f*TimePeriod4 + covariate1 + covariate2 + covariate3 + (1 | SBA),
data=data, REML= TRUE,
family=Gamma(link=identity)), control=glmerControl(
optimizer="bobyqa",optCtrl=list(maxfun=2e5)))
convergence code: 0
unable to evaluate scaled gradient
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
Warning messages:
1: extra argument(s) ‘REML’ disregarded
2: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
unable to evaluate scaled gradient
3: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
4: In summary.merMod(glmer(sba_nyc_fe_c ~ 1 + Gcode.f * T + Gcode.f * :
additional arguments ignored
5: In vcov.merMod(object, use.hessian = use.hessian) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
6: In vcov.merMod(object, correlation = correlation, sigm = sig) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
下一步尝试使用优化器L-BFGS-B：
summary(glmer(countOutcome ~ 1 + Exposure.f*TimePeriod1 + Exposure.f*TimePeriod2 + Exposure.f*TimePeriod3 + Exposure.f*TimePeriod4 + covariate1 + covariate2 + covariate3 + (1 | SBA),
data=data, REML= TRUE,
family=Gamma(link=identity)), control=glmerControl(
control = lmerControl(optimizer ='L-BFGS-B')))
convergence code: 0
unable to evaluate scaled gradient
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
Warning messages:
1: extra argument(s) ‘REML’ disregarded
2: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
unable to evaluate scaled gradient
3: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
4: In summary.merMod(glmer(sba_nyc_fe_c ~ 1 + Gcode.f * T + Gcode.f * :
additional arguments ignored
5: In vcov.merMod(object, use.hessian = use.hessian) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
6: In vcov.merMod(object, correlation = correlation, sigm = sig) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back var-cov estimated from RX
下一步尝试使用优化器nlminb：
summary(glmer(countOutcome ~ 1 + Exposure.f*TimePeriod1 + Exposure.f*TimePeriod2 + Exposure.f*TimePeriod3 + Exposure.f*TimePeriod4 + covariate1 + covariate2 + covariate3 + (1 | SBA),
data=data, REML= TRUE,
family=Gamma(link=identity)), control=glmerControl(
optimizer ='optimx', optCtrl=list(method='nlminb')))
convergence code: 0
unable to evaluate scaled gradient
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
Warning messages:
1: extra argument(s) ‘REML’ disregarded
2: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
unable to evaluate scaled gradient
3: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, :
Model failed to converge: degenerate Hessian with 3 negative eigenvalues
4: In summary.merMod(glmer(sba_nyc_fe_c ~ 1 + Gcode.f * T + Gcode.f * :
additional arguments ignored
5: In vcov.merMod(object, use.hessian = use.hessian) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
6: In vcov.merMod(object, correlation = correlation, sigm = sig) :
variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX
我缺少什么吗？为什么这个模型不能收敛？谢谢