Page 1 of 1

Like Stats and Bestfit

Posted: February 19 2018
by Nishanth Sasankan
Hi
I am using the GetDistGUI and plotting a variable that I introduced. Shouldn't the bestfit for the variable be where the likelihood is maximum? The reason I ask is because, when i click on the like stat from the data tab, it gives me a different value away from the peak of the likelihood function.

Any insight would be helpful.

Thanks
Nishanth

Re: Like Stats and Bestfit

Posted: February 19 2018
by Antony Lewis
The best fit only coincides with the peak of the marginalized probability if the full distribution is Gaussian (or more generally symmetric).

Note likestats is not the best fit - it's the best-fit sample (in high dimensions typically some way from the best fit).