STAMPS Larry 4 October 2024

This is the extention of the maximum likelihood estimator when you have less information. eg: when fitting for data for signal and background. If S(x) and B(x) are known, then MLE is the best. However if you dont know B(x) but only some condition on B(x) eg: smoothly falling. Then this is the optimal estimator.

Alternatives would be:

• f-tests to determine the order of the Bkg (need family of functions)
• Assume Try various families of functions ala H→γγ

Semi parametric inference

Is there any place for semi-parametric

ch23 of vandegard

"Efficeint influence function"- finding this is an art.

"score functions"

"Well, …now I know why we dont use semi-parameteric inference in physics"

"Gataux derivative"