I am getting logLikelihood as -1500 for one of the logistic regression analysis.
What does it mean statistically?
Asked
Active
Viewed 2,802 times
0
-
2It means that's the loglikelihood value – Dason Apr 08 '19 at 01:32
-
Does it mean it’s good or bad? I’m worried that it’s -1500. I’m not a stats person. I needed to learn from stats experts interpretation in layman’s term. – Apr 08 '19 at 01:42
-
Possible duplicate of: https://stats.stackexchange.com/questions/112451/maximum-likelihood-estimation-mle-in-layman-terms – Marius Apr 08 '19 at 01:56
-
1Also, there is no common scale for loglikelihood values, so we can't say if -1500 is good or bad in absolute terms. If you fit a few different models to the same dataset, you can compare their values. – Marius Apr 08 '19 at 01:57
-
Possible duplicate of [Maximum Likelihood Estimation (MLE) in layman terms](https://stats.stackexchange.com/questions/112451/maximum-likelihood-estimation-mle-in-layman-terms) – Sycorax Apr 08 '19 at 04:38
1 Answers
0
Log likelihood is just the log of the likelihood. You can read details of this (at various levels of sophistication) in books on logistic regression.
But the value, by itself, means nothing in a practical sense. You can't say if it is good or bad or high or low and changing the scale (e.g. moving from inches to cm) will change the loglikelihood.
Mostly, it is used when comparing models. The difference between two log likelihoods (on the same data) does have meaning. It is an indicator of how much better one model fits than another, and it is used in a lot of ways which, again, you can read about in books on logistic regression.
Peter Flom
- 94,055
- 35
- 143
- 276
-
2The question begs the question of whether basic reading should be undertaken before using a statistical method. – Frank Harrell Apr 08 '19 at 12:58
-
1