How to perform MLE for Censored and Truncated Data?

jonathanjka — Thu, 26 Jul 2018 19:41:01 GMT

Hi there, I'm new to the forum so feel free to redirect if I should post under a different location.

I understand that the following procedure can be done using PROC SEVERITY, but my company doesn't have a SAS/ETS license, so it would be appreciated if any suggestions can be limited to SAS Studio or Base SAS procedures.

I have a simple dataset from an insurance company, containing the following claims information:

Policy Deductible
Policy Limit
Incurred Loss Amount

I would like to fit this data to a few relatively standard distributions (eg. Beta, Gamma, Pareto, Lognormal, Weibull) using Maximum Likelihood Estimation, then compare the fit using Likelihood Ratio Test, SBC, AIC.

The problem is that to my knowledge, PROC UNIVARIATE doesn't easily allow parameter estimation for left truncated (by policy deductible) and right censored (by policy limit) data that I have.

Can someone help point me in the direction of how this could be accomplished?

Thanks in advance.

Re: How to perform MLE for Censored and Truncated Data?

Rick_SAS — Thu, 26 Jul 2018 19:47:51 GMT

You can program the log-likelihood functions in PROC NLMIXED. Some are distributions you mention are already built into PROC NLMIXED.

topic How to perform MLE for Censored and Truncated Data? in Statistical Procedures

How to perform MLE for Censored and Truncated Data?

Re: How to perform MLE for Censored and Truncated Data?