I'm new to the world of statistical modelling and have recently have been given a model to update. The model currently assigns a propensity score to each of the records in the database unless the record has a variable missing. If the record has a variable missing and a propensity score can't be appended then it assigns a percentage to the record based on a postcode level of geography (in this case the % of households in the postcode that have health insurance). My question is, can a propensity score and a % be directly compared or should there be some sort of weighting involved to make them comparable?
Advertise on Talk Stats