[R] fitdistr: was Heteroscedasity...

Collin Lynch collinl at cs.pitt.edu
Mon Oct 28 21:09:42 CET 2013

Hello again, first off thank you for your suggestion Mr. Rigby, I'll take
a look at the GAMLSS package.

I have a (slightly) related followup question regarding the 'fitdistr'
function.  I was examining my data, a sample of which is attached, using
this function and I am confused about the interpretation of the loglik

Based upon my experience I understand log-likelyhood value returned from
this function should be in the range -Inf - 0 with values closer to 0
being the best choice.  However when I test this data I get the following

fitdistr(E, "normal")$loglik
[1] 11.15125

fitdistr(E + 1, "lognormal")$loglik
[1] -0.8575117

fitdistr(E, "exponential")$loglik
[1] -73.18107

Is this a sign of error, in my system or on my part?  And if not am I
correct in interpreting lognormal as the best choice?

	Thank you in advance,
	Collin Lynch.

More information about the R-help mailing list