search

USEPA / CompTox-ToxCast-tcplFit2

A Concentration-Response Modeling Utility
https://usepa.github.io/CompTox-ToxCast-tcplFit2/
Other
1 stars 0 forks source link

Poly2 top reporting #93

Open kpaulfriedman opened 5 months ago

kpaulfriedman commented 5 months ago

Poly2 parabolic responses that look like an inverted U shape capture the model top as the top of the data points and not the top of the model curve. Is this intentional? I can’t remember from our discussions on the new poly2. Here’s an example.

tcplConf(user=', pass='', db='invitrodb', drvr='MySQL', host='ccte-mysql-res.epa.gov') tcplPlot(output='console',fld='m4id', val=7896464)

You can see other models would have similar tops to what is actually given, but the poly2 has this “tail” up that doesn’t get captured by the top. It’s not bad to just have the top be the top of the responses – but I wondered overall for tcpl what makes sense here to be consistent with how we fit and call the theoretical top.

Kelly-Carstens-EPA commented 1 month ago

Adding to poly2 investigations, I noticed a trend where poly2 did not fit to the lowest concentrations when the concentrations were near the same effect size as the highest concentration tested. The same fitting occurred in both directions (flipped the data by multiplying by -1). In these 2 example plots, bmed = 0.

image image