Tagging @twvalente for a potential model of adoption/disadoption:
$$
P(adopt_i) = \text{logit}^{-1}(\beta_{adopt} exposure_i + \beta^{adopt} x)
$$
$$
P(disadopt_i) = \text{logit}^{-1}(\beta_{disadopt} exposure_i + \beta^{disadopt} x)
$$
We could also make disadoption to happen randomly.
Originally posted by @gvegayon in #46 (comment)