Loglikelihood calculation for fitnlm

Hello,

Can you provide the formula for how loglikelihood is calculated when using fitnlm function with bisquare weights? I would like to replicate the value that Matlab outputs to fully understand how its calculated.

[Torsten]

Can you provide a link to an example where fitnlm is used for distribution fitting (loglikelihood) ? I thought "fitdist" or "mle" are the correct tools to do this.

[Thomas Hemming]

Our case is just a time series where we want to fit either a constant or a constant plus a coefficient with bisquare weights. fitnlm returns a loglikelihood value so it must be calculated somewhow, although I follow what you're saying (and kind of also why I'm asking how it's calculated in this case).

[Torsten]

Can you provide the formula for how loglikelihood is calculated when using fitnlm function with bisquare weights?

If it's not part of the MATLAB documentation, I'd contact Technical Support for the relevant literature source:

https://uk.mathworks.com/support/contact_us.html

Answer 1

[Sulaymon Eshkabilov]

See this example of MATLAB help documentation for nonlinear logistic regression.

[Thomas Hemming]

Thanks. I don't really see how that addresses my question.

[Sulaymon Eshkabilov]

Don't you see how fitnlm() fcn can be employed to find maximum likelihood explained in details in this help DOC for nonlinear regression model. Your question was "Can you provide the formula for how loglikelihood is calculated when using fitnlm function with bisquare weights?"

[Torsten]

We can see how it is used, but not how it is calculated internally.