Reevaluate mc6 model directionality flag

[madison-feshuk]

Review current logic for the model directionality questionable flag: https://github.com/USEPA/CompTox-ToxCast-tcpl/blob/dev/R/mc6_mthds.R#L78

Inactives are getting flagged due to 1 data point surpassing cutoff. Ex:

Consider updated logic to 1) filter so flag only applies to actives (where active is defined as absolute value of hitc >= 0.9) AND/OR 2) require minimum 2 data points surpassing cutoff in opposite direction of winning model top

[Kelly-Carstens-EPA]

The flag description currently describes: "Flag series if model directionality is questionable, i.e. if the winning model direction was opposite, more responses (resp) would have exceeded the cutoff (coff). If loss was winning directionality (top < 0), flag if count(resp < -1coff) < 2count(resp > coff). If gain was winning directionality (top > 0), flag if count(resp > coff) < 2count(resp < -1coff)."

The mc6 methods code does not reflect this definition. However, I think this is the intent of the flag and we should update the code to consider only responses that exceed cutoff in the 'opposite' direction (opposite sign from the top of the winning model). Presently, the code considers all response in the opposite direction, regardless of whether they exceed cutoff.

[ElizabethGilson]

Out of endpoints that are a hit and have the model directionality flag the majority have a winning model of poly2. The percent break-down is as follows

• poly2: 71.0%
• exp2: 16.1%
• exp5: 6.45%
• poly1: 6.45%

Here are some examples plots of endpoints with the flag and with a winning model of poly2:

[Kelly-Carstens-EPA]

Interesting examples. I have a couple questions about these endpoints. 1) Is this a biologically relevant gain-loss behavior? 2) If not, is this flag flagging something concerning about model fit? It seems like a good fit. I think this is flagged because the top of the model is positive (or maybe the it's negative if defined by max conc tested) and the majority of the responses are in the negative direction? It might be helpful to look into the sign of the top for these two.