0
$\begingroup$

I'm a little bit new to statistics, so I'm not sure what I really need.

I have a table like the following with some categorical/nominal values (like Gender and Age Group) and a ratio scaled value (DropOut, which is the week after registration in which a person dropped out of the program [so it's never zero or less] and it is not the Calendar Week):

enter image description here

I want to analyze which AgeGroups and Genders (and maybe more categorical values) are most likely to drop out early.

I want to solve this problem in R. Maybe you can give me some hints how to do this or you already have a concrete approach. :)

Thanks.

$\endgroup$
  • $\begingroup$ I believe the data you posted above does not have age and gender information of students who didn't leave the course, to the contrary it only has data of only those students who left the course? If yes, do you have age and gender data of students who didn't drop out from the course? your response to these 2 questions will give me direction for best possible solution. $\endgroup$ – Enthusiast Feb 12 '16 at 13:16
  • $\begingroup$ This is a good point you mentioned and you're right, the above table only contains students who left the course. But actually I have all the data and can aggregate them however it's best for the analysis. So I also have the age and gender data of students, who didn't drop out AND how much weeks they are already in the course (without dropping out). @MdAzimulHaque $\endgroup$ – ScientiaEtVeritas Feb 12 '16 at 13:41
  • $\begingroup$ Why not ordinal regression with DropOut as response? $\endgroup$ – kjetil b halvorsen Mar 12 at 9:34
0
$\begingroup$

Objective: To understand profile of the students who have already left: Descriptive analysis for the data the way you have posted. i.e. only those students who have left the course. You can create bar graph of gender for percentage of male and female who left the course and similarly another bar graph for age bands. I think excel should be good enough for this kind of work.

In my opinion instead of multiple correspondence analysis, you can try correspondence analysis of gender Vs. binary variable of those who left and stayed. One more correspondence analysis plot for age group. If you still want to do. here is a tutorial: https://www.youtube.com/watch?v=reG8Y9ZgcaQ

Objective: To find out profile of students are prone to leave the course: create a data set of all students including those who have not left. this should have 3 columns age group, gender and a binary coded variable. Also, binary variable should have 1 if student has left and 0 if student is still with us. Here is a tutorial on how to do this kind of analysis in R. I also believe that these two variables might not give you everything in terms of insights. there might be other factors such as past academic grades, annual household income number of members in family etc.

On the ethical side, i hope you are not building this kind of model for gender or age discrimination. Good luck!!

$\endgroup$

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.