topic Re: fit nonlinear regression MODEL in New SAS User

fit nonlinear regression MODEL

Joa14 — Thu, 28 Nov 2019 09:46:51 GMT

I am struggling with getting the right code on xls. exported data to estimate the initial values of model parameters (logistic, Gompertz, Richards, and Weibull) my ymax is 61.0

I tried with these two but I get error:

Any suggestion? Thanks!

DATA=ALL

YMAX=61.0

LOGISTIC=LOG((YMAX/Y)-1);

GOMPERTZ=LOG(-LOG(Y/YMAX));

RICHARDS=LOG(1-(Y/YMAX));

WEIBULL=LOG((-LOG(1-(Y/YMAX));

LNX=LOG(X);

RUN;

PROC REG DATA=ALL

MODEL LOGISTIC=X;

MODEL GOMPERTZ=X;

MODEL RICHARDS=X

MODEL WEIBULL=LNX;

RUN;

Re: fit nonlinear regression MODEL

PaigeMiller — Thu, 28 Nov 2019 14:57:12 GMT

There is no equal sign in a DATA statement. DATA statements must end with a semi-colon. Assigning values must end with a semi-colon. Variables Y and X are never assigned a value. The number of open parentheses in a statement must equal the number of close parentheses in the statement.

Also you can't fit a regression to this data, it has only one data point.

Re: fit nonlinear regression MODEL

PGStats — Thu, 28 Nov 2019 18:51:50 GMT

Syntax should be:

DATA ALL;
YMAX=61.0;
set mydata; /* dataset mydata contains X and Y variables */
LOGISTIC=LOG((YMAX/Y)-1);
GOMPERTZ=LOG(-LOG(Y/YMAX));
RICHARDS=LOG(1-(Y/YMAX));
WEIBULL=LOG(-LOG(1-(Y/YMAX)));
LNX=LOG(X);
RUN;

PROC REG DATA=ALL;
var X LNX LOGISTIC GOMPERTZ RICHARDS WEIBULL;
Logistic: MODEL LOGISTIC=X;
Gompertz: MODEL GOMPERTZ=X;
Richards: MODEL RICHARDS=X;
Weibull: MODEL WEIBULL=LNX;
RUN;

(untested)

Re: fit nonlinear regression MODEL

PaigeMiller — Thu, 28 Nov 2019 18:04:28 GMT

Still needs to have open and close parentheses match.

Re: fit nonlinear regression MODEL

PGStats — Thu, 28 Nov 2019 18:53:04 GMT

Thanks @PaigeMiller . I edited the code.

Re: fit nonlinear regression MODEL

Joa14 — Fri, 29 Nov 2019 00:35:18 GMT

Thanks, PG. It is still not working but will keep trying.

Re: fit nonlinear regression MODEL

Joa14 — Fri, 29 Nov 2019 00:35:41 GMT

Thanks, PaigeMiller!

Re: fit nonlinear regression MODEL

PGStats — Fri, 29 Nov 2019 04:38:57 GMT

We would need to see the SAS log (the part showing the code and warnings/errors) to help you further.

Re: fit nonlinear regression MODEL

Joa14 — Sat, 30 Nov 2019 03:08:29 GMT

100 PROC REG DATA=WORK.IMPORT;

101 var X LNX LOGISTIC GOMPERTZ RICHARDS WEIBULL;

102 Logistic: MODEL LOGISTIC=X;

103 Gompertz: MODEL GOMPERTZ=X;

104 Richards: MODEL RICHARDS=X;

105 Weibull: MODEL WEIBULL=LNX;

106 RUN;

ERROR: No valid observations are found.

Re: fit nonlinear regression MODEL

PGStats — Sat, 30 Nov 2019 03:37:31 GMT

This message usually signals that all observations contain missing values. Please check the contents of dataset WORK.IMPORT.

Re: fit nonlinear regression MODEL

Joa14 — Sat, 30 Nov 2019 03:49:05 GMT

It seems ok to me. The data xls. format is attached if you could check. It has DBH (X) and HT (Y).

I want to do this because my goal is to use the modified logistic functions to fit the function to the data using three iteration methods available in PROC NLIN (GAUSS, MARQUARDT, and NEWTON). Then compare the parameter estimates, asymptotic standard errors for the parameters, and asymptotic correlations among the parameters for the three methods.

Re: fit nonlinear regression MODEL

PaigeMiller — Sat, 30 Nov 2019 14:59:15 GMT

When @PGStats says check the data, he specifically means (and specifically stated) check the SAS data set you are using, specifically you need to look at WORK.IMPORT and see if everything is okay in there. The XLS file is irrelevant.