From Tom (via a conversation with David Krackhardt).
There's a paper showing that the variability of stress levels in one's network was associated with individual stress. It would be nice to have a toggle to compute the standard deviation (and maybe other moments?) of the outcome when calculating exposure, in addition to the proportion or count.
This should work for both continuous and binary data even when it can get a bit funky at low degrees, but we can get by possibly with some bootstrapping.
Re-analyzing the classic datasets using variance as an additional control could be worth.
From Tom (via a conversation with David Krackhardt).
There's a paper showing that the variability of stress levels in one's network was associated with individual stress. It would be nice to have a toggle to compute the standard deviation (and maybe other moments?) of the outcome when calculating exposure, in addition to the proportion or count.
This should work for both continuous and binary data even when it can get a bit funky at low degrees, but we can get by possibly with some bootstrapping.
Re-analyzing the classic datasets using variance as an additional control could be worth.