Handling cases where main slice has differing number of metrics than its CI

[flemingjf]

Since Toxpi sums metrics, if a CI has more or less metrics than the main score the unscaled slice score for the CI could be inflated or deflated

[flemingjf]

Proposal for handling: Alter the ToxPi framework to use averages instead of sums. Since all samples contain the same number of metric columns for a given slice (n is fixed between samples for a given slice), taking the average or sum (assuming NA = 0) will result in the same slice score after rescaling 0-1, thus the new method will be consistent with the old in determining slice scores.
By using the average, if n is different for the main slice and one of the CI interval slices, inflation or deflation won't occur allowing the aggregated main slice unscaled scores and aggregated CI slice unscaled scores to be directly compared in magnitude and rescaled 0-1 together