Practical Predictive Analytics

Adjusting records to simulate an intervention

Now that we have multiple records, we will be able to adjust for the new survey information which changed after month 6 and include that as a time dependent variable.

Recall that we initially simulated the second survey data for the churners by increasing the satisfaction rating by 1 (Xsatisfaction2). This resulted in some satisfaction scores of 6 for some members, which would be impossible. So we will first clean those by setting the 6 ratings to 5:

#fix up some "6" satisfaction scores, that are not possible, and make them "5"SURV2$Xsatisfaction2 <- as.factor(ifelse(SURV2$Xsatisfaction2=="6","5",SURV2$Xsatisfaction2))

Assume that at month 5 there was a promotion targeted to those members who ...