I have a large data set with multiple records per phone number. Each record has two variables - the number of past attempts in the past 60 days (number of times it was called) which range from 0 to 30 or so and a success indicator (0 or 1). I want to model how the number of past attempts impacts the probability of the outcome. There are no other covariates.
I want to model it as outcome~pastAttempts. The problem is that the records are not independent, as they by definition are the same phone numbers. Some phone numbers are listed 1 time, others up to 98 times. Can I use logistic regression for this since the covariate pastAttempts is adjusting for the fact there are repeated observations? If not, what other technique is plausible?
The data set covers a time period greater than 60 days and the predictor variable is the number of attempts in the past 60 days. So it is not always increasing but can "reset" along the way since the variable is a "rolling" measure.
here is an example of 4 phone numbers.