The Log L (log likelihood) in my FIML printout is positive, but I thought that probabilities were between zero and one, so that their logs were negative?
It’s true that probabilities are between zero and one, but the FIML likelihood function is a density (as opposed to, say, a cdf like in PROBIT). A density can take on any positive value, so its log can be positive (or negative, or zero). For example, the uniform density on the interval [0,.5] has the value 2 at all points. The normal density for a zero residual, with mean=0 and variance=(1/(25*2*pi)), is 5. So a normal density will be largest when the variance and residual are fairly small, i.e. the peak of the “bell curve” can be quite tall when it is narrowed by a small variance. Finally, the FIML likelihood function (in particular) includes the determinant of the Jacobian, which can also be larger than 1.
Related Questions
- I forgot my Keylogger password so I can not open the Log Viewer to view the logs. Can You, Supremtec, access the logs on my computer and send them to me?
- Prior to 11.0 debug messages and other Illuminator log messages could be found in the ServletExec logs. Where are these messages written in 11.0?
- What are some adjectives that have a positive conotation and start with the letter l?
