Polishing of CDI Utilities 1

[adriansteffan]

I have added a first version of utilities to deal with cdi data, but there are some areas where they could use polishing:

A) Validation of CDI data

• there now exists the function cleanup_cdi_data that is supposed to catch anomalies that make analysis difficult. For now, it only catches duplicate cdi values for the same administration (we need a validator for this anyway - see Find the participant with duplicate cdi data peekbank-data-import#146) - we should brainstorm and implement other anomalies that we would like to catch here

B) Relative CDI scores

• I took some code from the peekbank method repo that calculates a percentage of points reached out of the maximum for a given cdi type. However, I could use the eyes of someone experienced with cdi to determine the correct comparison values for various languages and cdi types.

C) Percentile calculation for CDI rawscores (benchmarking)

• I have added a function that gets the respective cdi percentile for each rawscore loosely based on work done by george and the others https://github.com/kachergis/cdi-percentiles/tree/main. At the moment, this is hardcoded for American English and the 2022 norm, so we might want to add more languages and norms and make the function more flexible as a result. Additionally, this function could use some testing + a look by someone more experienced with cdis (open point: what to do with kids that are too young or old? right now, the function defaults to the nearest available age)

[mzettersten]

@mcfrank Adrian has been working on CDI score handling, including computing CDI percentiles. Looping you in here!