I am fully aware that question regrading left and right censoring has been asked before. I will however post my own question, as I believe that its focus differs significantly. Here goes:
I have a dataset with customer data where I would like to estimate a PH model describing the probability for churn. I have (aside form a bunch of covariates) data on when the customer starts and when the customer stop, BUT both might be censored. For some individuals there might even be censored in both ends. I have data from the period 2018-2019, so if the customer is already buying the first week of 2018, he is 'left' censored, and if the customer is still buying in the last week of 2019 he is 'right' censored. This is how I originally though of the data.
But as input to my estimation of a proportional hazards model, I'm using the length of the period that the customer is active. So if a customer is for example active for the last 5 weeks of 2019, I just know that he is at least active for 5 week, and if he is active for the first 5 weeks of 2018 i just know that he is at least active for 5 weeks. So either way the length of the period that the customer is active are 'right' in both cases.
I regard left censoring a lot harder to navigate, and therefore it would be lovely if i could regard all my variable as being right censored.