# How to use Correlation of variables to predict future pattern

I have data set of file generating from some user. User can generate 1, 2, 3,4, 5 and more files during a day. Data set example:

Day one - 2 files generated - first is at 09:10 and second one at 10:30

Day two - 1 files generated - first at 08:30

Day three - 3 files generated - first at 09:15 and second one 10:45 and third at 12:00

I managed to determine distribution for number of files, so now I can generate new data set that will "simulate" creating of new files, but how to determine when this files will be generated. Should I look for corelation between time of generation of files and number of files for some day. I believe that since one variable is discrete and other is not that I should do some transformation.

There is no influence number of files on time of creation. I done histogram for time of creation of only first file and all days with more then one file no big difference.

and all days / first file creation

If i want to generate similar time of creation how to do that, If this are counted data I would determine distribution and use this to generate numbers, but how to do this if I need to generate time?

• Could you explain what you mean by two variables? I see a description of only one: times of file generation. If you would like to simulate a new dataset, then you ultimately need a model for this stochastic point process. We cannot invent models for you, because a good choice will depend on your understanding or theory of the underlying process and on the actual data. – whuber Dec 14 '18 at 14:55
• One variable is number of files durring a day and another is time of creation. If i have more files per day then this will efect time of creation for first file, swcond and so on. Do you still think that this is one variable,mybee my starting point is wrong. – explorer Dec 14 '18 at 15:48
• The underlying variable is time of creation, but you are positing an (interesting) model in which that is related to the number of files per day (a statistic that is derived from the variable). You might be able to make progress if you could describe, in the most quantitative possible way, how you think the daily file counts affect the file creation times. – whuber Dec 14 '18 at 16:14
• if I take all days with only one file created data shows that creation will happened around 09:51 (average value). If I take all days with two files creation, first file will be created around 09:35 o clock but second file is created around 11:22. If I take all days with 3 files first creation will happened around 08:54, second file around 09:55 and third file around 11:43. and so on. I can conclude that if we have more files / day first file creation in average is earlier, second file is moving closer to first file and so on... – explorer Dec 14 '18 at 17:47