1
$\begingroup$

I'm conducting a Kaplan Meier study to determine user retention for a website. I'm planning to limit my timeline to 6 months and see how many users are dropping within that time period. And if a user does not have any action for a month I'm considering that as an event. But I keep seeing new users. What should I do in that case? Limit the users in my study to all the users I see in the first month?

$\endgroup$
  • 2
    $\begingroup$ I gather your drop-out event is inaction for a month. But what if a user comes back on-line a few weeks later? Do you consider that a new user? Or do you erase the drop-out event? So does "survival" mean some kind of user activity within the last 4 weeks of the end of your trial? You'll have to be explicit about how you handle dropping out and re-upping. You could also study the distribution of dwell times between visits to the site. If you fit the low end to a common distribution, you could have a data-defined "death event dwell time" (6 weeks? 8 weeks?) to use with the survival study. $\endgroup$ – Peter Leopold May 30 '19 at 15:40
  • 2
    $\begingroup$ At first, the the definition of beginning is needed to calculate the survival time for K-M method.For example, it can be the the time the user opens his account. $\endgroup$ – user158565 May 31 '19 at 3:42
5
$\begingroup$

Do not use calendar time to display the Kaplan Meier curve. Rather, take the time at which a user establishes an account, or first navigates to the site (or whatever data you have) as time 0.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.