Hi,
I am working on the Klein-Spady semiparametric estimator.
I am trying to use call NLPNRA to generate a maximum likelihood estimate. IML gives an ERROR message saying that I am taking log of zero. But I checked the arguments in all the log commands, none of them is zero.
Below is the program and the error message. Thanks for your help.
ERROR: (execution) Invalid argument to function.
operation : LOG at line 296 column 45
operands : _TEM1008
_TEM1008 1 row 1 col (numeric)
0
statement : ASSIGN at line 296 column 1
traceback : module LOGLIK at line 296 column 1
ERROR: NEWRAP Optimization cannot be completed.
ERROR: The function value of the objective function cannot be computed at the starting point.
I don't understand what you are doing well enough to diagnose what is going wrong, but from the error messages I assume that the LINK variable is getting out of the open interval (0,1). If you put in the line
link = choose( link<=0 | link>=1, ., link);
then the algorithm converge. Unless there is a theoretical reason why a quantity cannot be less than zero, it's a good programming practice to protect against taking log(negative_value).
First, look at the following article: Maximum likelihood estimation in SAS/IML - The DO Loop
What are the meanings of your three parameters (for example, location, scale, or shape parameters). Are any of your parameters constrained? If so, see Step 2 in the article.
thank you,
it's just a binary model, the parameters are not constraint and the initial values come from the proc logistic.
the module computes the log likelihood so i don't understand the problem
I don't understand what you are doing well enough to diagnose what is going wrong, but from the error messages I assume that the LINK variable is getting out of the open interval (0,1). If you put in the line
link = choose( link<=0 | link>=1, ., link);
then the algorithm converge. Unless there is a theoretical reason why a quantity cannot be less than zero, it's a good programming practice to protect against taking log(negative_value).
thanks for this advice, it's work fine
Don't miss out on SAS Innovate - Register now for the FREE Livestream!
Can't make it to Vegas? No problem! Watch our general sessions LIVE or on-demand starting April 17th. Hear from SAS execs, best-selling author Adam Grant, Hot Ones host Sean Evans, top tech journalist Kara Swisher, AI expert Cassie Kozyrkov, and the mind-blowing dance crew iLuminate! Plus, get access to over 20 breakout sessions.
Learn how to run multiple linear regression models with and without interactions, presented by SAS user Alex Chaplin.
Find more tutorials on the SAS Users YouTube channel.