[exekias]

How does using regularization affect inference? If your parameters aren’t MLE then the score (gradient of loglikelihood) isn’t zero. So, it seems like a lot of the assumptions behind the calculation of standard errors and p-values are thrown off. Does anyone have a good reference?

[exekias]

Yeah you don’t really need that much detail to get a significant advantage. Even just knowing x,y,z topics weren’t tested but a,b,c were will improve someone’s chances. They could then double their efforts on a,b,c and not waste time practicing x,y,z.

 

Those types of things are also more likely for someone to let slip to a friend too. “Oh man they had some really tough questions on b.” Or, “I’m surprised they didn’t cover x at all.” Still, I wouldn’t expect there to be enough people doing this to impact the pass mark. If you learn the material well enough, you will pass regardless of what other candidates are doing.

[exekias]

I used the CF/RF joint seminar and I feel like it prepared me very well. If you need to save money the CF seminar would probably be enough on its own. It does a good job of explaining all the material and the problems at the end of each section are really helpful for testing your knowledge.

[Author]

Posts