how to calculate the parameters using proc nlmixed with Standard-Exponential–distribution

Jie111 — Wed, 27 May 2020 13:40:21 GMT



proc nlmixed data=lab3; /*the dataset include two variable: ID, age for death*/
parms mu 1 sigma 1 ;   
bounds 0 < sigma; 

G_t=exp ((age- mu)/sigma);
g=exp((age-mu)/sigma)/sigma;
ll= exp(- G_t)* g;
model age~ general(ll);
run;

I wrote the code based on the below conditions:

We consider the location-scale-skew transform
G(y)=exp[( y−μ)/σ] with derivative g(y)=exp[( y−μ)/σ]/σ.

We assume that G(Y)∼fE(y)=exp(−y). Then Y follows a generalized extreme value distribution fSE(y)=fE[G(y)]g(y)=exp[−G(y)]g(y)

however, the code does not work well.

NOTE: Convergence criterion (FCONV=2.220446E-13) satisfied.
NOTE: At least one element of the gradient is greater than 1e-3.
NOTE: Moore-Penrose inverse is used in covariance matrix.
WARNING: The final Hessian matrix is full rank but has at least one negative eigenvalue. Second-order
         optimality condition violated.
NOTE: PROCEDURE NLMIXED used (Total process time):
      real time           1.98 seconds
      cpu time            1.61 seconds

does anyone know how to do this in SAS?

Re: how to calculate the parameters using proc nlmixed with Standard-Exponential–distribution

Jie111 — Thu, 28 May 2020 11:41:36 GMT

proc nlmixed data=lab3;
parms mu 80 sigma 5 ;
bounds 0 < sigma; 
ll= -exp((age - mu)/sigma) + (age - mu)/sigma - log (sigma);
model age~general(ll);
run;

updating.

The above syntax works well this time. it seems that I need to define the initial value for mu and sigma near the value of parameters. Is there any way to deal with the initial value or do I have to give the initial value artificially?

topic how to calculate the parameters using proc nlmixed with Standard-Exponential–distribution in SAS Programming

how to calculate the parameters using proc nlmixed with Standard-Exponential–distribution

Re: how to calculate the parameters using proc nlmixed with Standard-Exponential–distribution