Implement COMPS prediction option
Now that we have #97 (closed) completed and in alpha testing, we want to be able to leverage the comp-set to estimate property values. This is relatively straightforward. In data_dictionary_constituents, models contains a list of models we use. Add your comp method to the list an populate accordingly.
Each field is control used in 2_sf_modeling. If you have to create new values in the models CSV, do so and make sure you account for them in 2_sf_modeling. Create a new chunk of code in 2_sf_modeling so that, when the loop encounters the COMPS estimator, it knows what to do.
Second part. In 3_sf_valuation, create a new step between A & B (?I'm not sure where it is appropriate to include this?) where you compare the model output for each property to the set comps for that property. Think about what it means that a model result is significantly different from the comps result - does that create an opportunity for an appeal? If so, why not adjust the value ex ante, precluding the need for an appeal? Create a binary indicator that flags this.
In order to develop this idea, look back at the relationship between comps, first-pass values, and actual appeals. @SweatyHandshake knows where to get the appeals data from.