0
$\begingroup$

I have dataset about file creation durring a year. From this dataset I extract subset for days when user created just one file. there is information (timestamp) when that happened. I created a bins with 15 minutes time window and this are my data. Time reflect working hours

Time elapsed / frequency
0 -15 / 0
15 - 30 / 0
30 - 45 / 0
45 - 60 / 0
60 - 75 / 4
75 - 90/13
90 - 105/ 6
105 - 120 /3
120 - 135 /5
135 - 150 /3
150 - 165 /2
165 - 180 /6
180 - 195 /3
195 - 210 /2
210 - 225 /2
225 - 240 /2
240 - 255 /1
255 - 270 /1
270 - 285 /2
285 - 300 /2
300 - 315 /2
315 - 330 /1
330 - 345 /2
345 - 360 /0
360 - 375 /0
375 - 390 /2
390 - 405 /1
405 - 420 /0
420 - 435 /1
435 - 450 /1
450 - 465 /0
465 - 480 /0
480 - 495 /0
495 - 510 /0
510 - 525 /0
525 - 540 /0
540 - 555 /0

I need to determine probability when generation of file will happened. Based of graph the biggiest probabilty is from 75 - 90 minutes from 07:00 if we take this as starting point for workhing day.

enter image description here

Should I use logarithm distribution in order to produce pdf for generate time for creation of file.


I reduced time bins for 15 minutes, following recommendation i use all days and record first file creation and get this table and graph

my frequency table enter image description here

and graph looks like

enter image description here

$\endgroup$
  • $\begingroup$ Maybe. But before you even get to that, two things worry me about this data set. First, why do you take the subset of days when only one file is created? Maybe the distribution is different on days when several files are created, in which case this would probably be a biased sample, depending on what your goal is. On the other hand, if the distribution isn't different, then you're probably discarding perfectly good data (the time at which the first file is created on a day when multiple files are created). $\endgroup$ – The Laconic Dec 22 '17 at 12:58
  • $\begingroup$ Second, why are there no observations at less than 60 minutes? If it takes at least 60 minutes to create a file, that would make sense, but then an exponential (I think you mean exponential, not logarithmic?) distribution isn't appropriate. At least, not without shifting the distribution by 60 minutes (or whatever). $\endgroup$ – The Laconic Dec 22 '17 at 13:01
  • $\begingroup$ First 60 minutes are without files since working hours start at 08.00 but i measure one hour earlier since some one could come earlier :) but that s not that case. So you think thant I should observ all times of creation first file no matter what is number of files per day, i will do that. $\endgroup$ – explorer Dec 22 '17 at 16:59
  • $\begingroup$ @TheLaconic i add all times with first file creation and graph regarding this new dataset. $\endgroup$ – explorer Dec 24 '17 at 8:55
  • $\begingroup$ How to find mean and standard deviation from hours ? I believe that gamma distribution is possible solution for this. $\endgroup$ – explorer Dec 26 '17 at 9:27

Your Answer

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

Browse other questions tagged or ask your own question.