1
$\begingroup$

I have a set of game play event data. One of our goals is to provide a picture of how much time the person has spent playing. This sounds simple but for the fact that events are often dropped and result in vast amounts of time between events.

We are in field trials right now and don't have a huge pool of data to use as reference.

I don't have much experience in statics, though I'm a competent programmer. My question is: What might I do to account for the missing data? How do I ensure that these missing events don't skew my data?

$\endgroup$
1
$\begingroup$

What you have described is that you want to consider the distribution of observations in a way that models the population to serve a better purpose.

I would recommend reading into how different types of distributions can be transformed so that your model might achieve a higher score.

In terms of your plots during analysis of your dataset, I would look at resampling at different frequencies, this will change how your plots will look and allow you to identify issues with seasonality, such as breaks in the account activity.

If the observations are missing because the person simply wasn't active, then this is a good representation of reality, the question is maybe how would you increase the users engagement?

| 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.