z-scores

[palewire]

@ryanvmenezes, think that's something that could be boiled down to simple python in here for reuse in our other projects?

[ryanvmenezes]

@palewire definitely. I'll try to dummy up some code using existing functions. I think the two most useful would go something like this:

1. t-value for simple linear regression line of best fit
   tvalue_SLR(indepVarList, depVarList, nullSlope=0)
   null hypothesis: slope of best-fit line is 0 (no relationship between variables)
2. z-value for a single proportion
   zvalue_proportion(successList, failureList, mean=sum(successList)/sum(failureList))
   (this is what I'm ranking on for the overcrowding analysis)

[palewire]

Are there existing Python functions that do this already in numpy or scipy?

[ryanvmenezes]

@palewire Ah, scipy.stats.linregress is an existing version of my no. 1, though it doesn't return a t-value, which I think is most important. The p-value it returns can be reverse-engineered to be a t-value using scipy.stats.t in a custom function, I think.

Don't see a version of my no. 2 yet...

[palewire]

Hmm. It might be a fun exercise to unpack the scipy function and see how we could rewrite it in a simplified, more literate and documented way.

[ryanvmenezes]

@palewire I just recalled this request because I've been writing a lot of python scripts that don't quite generate the z-/t-scores I want. Would it be against the spirit of this repo to import a package like scipy or numpy? My guess is yes but I'll let you decide. In any case, function no. 2 (the proportion z-value) is doable with simple arithmetic.

[palewire]

What do you need from numpy to make your function work?