I have a data set in which I have the date of joining of employees, age, and the date of leaving , also i have the dataset with current employees,so should I combine these two datasets or should I carry on with my analysis of already left employees, also i still dont understand properly time variable and status variable. if i can find out attrition based on gender of the employees, i have have their performance ratings as well. If you can please help me out with this, i will be highly obliged.



  • $\begingroup$ Although time is an essential part of any survival analysis, "status" is not. Would you mind clarifying for us what you mean by "status variable"? $\endgroup$ – whuber Jun 19 '15 at 12:40
  • $\begingroup$ Hello!! By status I mean the event variable , like if the company has adopted a measure for controlling attrition, then number of employees who left the company after 6 months of that measure. $\endgroup$ – Neal Jun 22 '15 at 4:49

By your description I believe that you are trying to estimate the probability of a employee leave/stay the company after, eg, 6 months, 1 year, and so on.

So, all your employees will count, actives and non-actives. For survival analysis you need 2 variables: one referring time (quantitative) and the other referring the status active/non-active (binary variable).

For each employee, day 0 is when they start working in the company. If an employee leaves after 6 months (180 days), then you mark it as non-active. Otherwise the employee is considered active and you have to censor them at the last known day of work (today's date - start date).

I usually do survival analysis using days as a unit. Maybe for your is better using months.

After this first analysis you can start stratifying by gender and study if both survival distributions have the same behavior (you can use a Logrank test).

  • $\begingroup$ Thanks Yohana I have a clear understanding now. I peformed a survival analysis in Rattle, I am getting a concordance of .67 , R square of .04, concordance I understand but what about the R square value, as it is very low. Also if you can tell me a book I should read , or a better software for doing it, Thanks a lot for your help $\endgroup$ – Neal Jun 22 '15 at 4:55
  • $\begingroup$ You are welcome! Well, you will see that all software gives you a lots of measures. About the R square take a quick look in apha.confex.com/apha/134am/techprogram/paper_135906.htm. Books, for survival I have "a bible" which is extremely big "Survival Analysis - Kleinbaum" springer.com/us/book/9781441966452. But I would suggest you to take a look to these 2 dictionaries "The Concise Encyclopedia of Statistics - Yadolah Dodge" and "THE CAMBRIDGE DICTIONARY OF Statistics - B. S. EVERITT". $\endgroup$ – yohanna Jun 23 '15 at 6:47

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.