log likelihood

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

dbyroo1
Posts: 3
Joined: Thu Apr 26, 2012 5:42 am

log likelihood

Postby dbyroo1 » Wed May 02, 2012 2:56 am

Can anyone tell me how to interpret the log likelihood. I know they are supposed to be used to compare which model fits your data better.
I have estimates of:
Regression 1 -222.73
Regression 2 -223.54
Regression 3 -228.31

I want to comment on the models to justify dropping and including variables but I'm not sure if it is good to have a higher or lower negative number!

EViews Esther
EViews Developer
Posts: 149
Joined: Fri Sep 03, 2010 7:57 am

Re: log likelihood

Postby EViews Esther » Wed May 02, 2012 8:59 am

When a model has more parameters, it will fit the model better. Therefore, it is required to balance how good you fit the data with a variety of models.

It is common statistical practice that likelihood functions are used in the comparison of models for the same data. However, I do not recommend to use the loglikelihoods in your model comparison.

There have been several proposals to address this issue.
1. EViews also provides the Akaike Information Criterion (AIC) or Schwarz Criterion (or BIC). The best model usually displays the lowest AIC/BIC.
2. The BMA addin provides the most likely models as well as the averaged quantities of interest over all possible models.

Khatai
Posts: 4
Joined: Fri Apr 18, 2014 4:28 am

Re: log likelihood

Postby Khatai » Wed Apr 30, 2014 10:26 pm

İf Akaike Information Criterion (AIC) or Schwarz Criterion (or BIC) values are negative, we must choose the lowest value in absolute value to define the best fitted model?

EViews Esther
EViews Developer
Posts: 149
Joined: Fri Sep 03, 2010 7:57 am

Re: log likelihood

Postby EViews Esther » Thu May 01, 2014 8:25 am

We will choose the model with the smallest value.


Return to “Econometric Discussions”

Who is online

Users browsing this forum: Google [Bot] and 3 guests