相关疾病： 用nlximed运行多水平零膨胀负二项模型，用多水平logistic、多水平负二项做logit和负二项部分初始值，得出结果好多的".",日志有错误报告：WARNING: The final Hessian matrix is not positive definite, and therefore the estimated covariance matrix is not full rank and may be unreliable. The variance of some parameter estimates is zero or some parameters are linearly related to other parameters. 该怎么修改程序？是哪里出的问题？初始值设定多水平的有没有问题？ 程序：proc nlmixed data=sasuser.c7;/*多水平零膨胀负二项(logistic yn)*/ parms A0=-0.2285 Asex=-0.5112 Aage=-0.00767 Aedu=0.07857 Aq32=0.2301 Avolun1=-0.2291 Adrink=-0.4512 Ac10new=0.2190 Ac11new=-0.2265 Apress=-0.1642 B0=-0.09054 bsex=-0.4259 bage=-0.00888 bvolun1=-0.2152 bc10new=0.1656 bc11new=-0.2021 v_u1=1,cov_u12=1,v_u2=1 k=1.85; s1=A0+Asex*sex+Aage*age+Aedu*edu+Aq32*q32+Avolun1*volun1+ Adrink*drink+Ac10new*c10new+Ac11new*c11new+Apress*press+u1; p=1/(1+exp(-s1)); s2=B0+Bsex*sex+Bage*age+Bvolun1*volun1+Bc10new*c10new+Bc11new*c11new+u2; lambda=exp(s2); pdf=gamma(n+1/k)/(gamma(n+1)*(gamma(1/k)))*(((k*lambda)**n)/(1+k*lambda)**(n+1/k)); if n=0 then pdf=p+(1-p)*pdf; else pdf=(1-p)*pdf; if pdf>1e-8 then ll=log(pdf); else ll=-1e20; model n~general(ll); random u1 u2~normal([0,0],[v_u1,cov_u12,v_u2]) subject=site; run; 程序结果（部分截图）：