Quote:
Originally Posted by Unregistered
Knowing how to click on statistical software is different from knowing how to use regression analysis.
Did u do any diagnostic tests to check for normality, multicollinearity, and homoscedasticity?
if the model failed any of the test, how do u proceed?
How do u choose the the independent variables? by experience? based on empirical result? stepwise or some other forms of selection process? Is the model variation in discrete selection process a concern? If so, should I use penalty regression instead?
And that is just the model building stage. There are many more other issues to take note of at the inference stage.
|
Yea we do have some data QC procedures that we need to check for before processing. Normality yes, the other 2 I don't know.
I'm not trying to offend anybody who truly likes data crunching or doing academic stuff, but I highly doubt most normal jobs out there in the consultancy world really require the sort of knowledge you are saying. These sound more like university research than real world consultancy.
I look around my office of 70+ people and maybe 2-3 principals who are actuaries will be involve in this sort of stuff, but then they are very experienced subject matter experts and are probably paid like >20k. The rest of us are really just doing simple analytic. I am sure Neilsen isn't hiring him at 4k to do any of the things you are suggesting.