Thread: MLE for non-differentiable functions

Hi,

I'm studying for a stats exam and have many questions. Here's the first.

When finding the MLE for non-differentiable functions, you write down the pdf, then the likelihood function L(u), then, if that's a different value depending on the value of the parameter of interest (u), you use a graphical method.

Do you simply draw the graph and show that L(u) is highest at some value of x? And that is your MLE?

I am also wondering why the Exponential distribution is undifferentiable, but the Poisson is not? The Poisson is also only for values of u>0.

Please explain! Ta,

-Canger

I knew someone on the web would have mentioned MLE for exponential distributions already. Enjoy.

what I'm assuming you mean, is that the support involves the parameter.
In those cases the largest or smallest order stats come into play