CODEWITHSUNDEEP
rstatisticsprobabilityprojectionsurvival-analysis
VASISTA
VASISTA

Reputation: 75

R: Creating probability/projection each month conditioned on survival probabilities

I have a data by ID by month, and for each month I have a customer churn probability something like below.

How can I get a churn probability each month conditioned on survival probability of previous month? For example for ID 6800, in the month ending 4/30 I have 0.54 probability, so survival will be (1-0.54)=0.46. Hence for month 5/31 I would like probability as 0.46*0.125. For month 6/30 I should have (1-0.46*0.125)*0.081440443. I want to do this for every ID for many months.

Is there a automated way I can set this up in a package or something?

    ID        Date     ChurProb     cash
1 6800  04/30/2017  0.541440443 35019.86
2 6800  05/31/2017  0.125103719 35047.55
3 6800  06/30/2017  0.081440443 15124.00
4 6801  04/30/2017  0.541440443 35019.86
5 6801  05/31/2017  0.125103719 35047.55
6 6801  06/30/2017  0.081440443 15124.00

Upvotes: 1

Views: 54

Answers (1)

user295691
user295691

Reputation: 7248

I looked for a more elegant solution, but came up empty. Here's a loop that gets the job done:

calc_cond_churn <- function (dta) {
  last_id <- -1
  last_survival <- 1
  CondChurn <- rep(0, nrow(dta))
  for (i in 1:nrow(dta)) {
    if (dta$ID[i] != last_id) {
      last_survival <- 1
      last_id <- dta$ID[i]
    }
    CondChurn[i] <- dta$ChurProb[i] * last_survival
    last_survival <- 1 - CondChurn[i]
  }

  CondChurn
}

You would use it as:

df$ConditionalChurn <- calc_cond_churn(df)

Upvotes: 1

Related Questions