You need to better frame this. So what are your assumptions, say - more than a random amount of people switched groups as if there actual movement was dictated by a fair coin flip?
Hello folks,
I am working with categorization of subjects in two different periods, and I want to test whether people migrated from categories or if they stick in the same category. My hypothesis is something like:
The participants change the group after interacting for 6 months
I my settings, I check to which group each subject is part during the first six months and categorize them. In the next period of six months, I categorize them again. So, I will have something like:
Subject;cat_before;cat_after
P1;A;A
P2;A;B
P3;C;C
P4;B;B
P5;D;C
P6;D;D
P7;B;B
I want to know, which kind of test should I use, since I am dealing with paired, nominal data.
Any clues?
You need to better frame this. So what are your assumptions, say - more than a random amount of people switched groups as if there actual movement was dictated by a fair coin flip?
Stop cowardice, ban guns!
Thanks for your answer hlsmith.
So, I am analyzing an online community and want to verify if people become specialists in something after sometime. To do so, we analyze in which kind of projects this people are working on and classify them according to the projects they work (each subject is classified as specialist in ONLY one thing). We perform such classification in two different periods and we want to check whether people tend to work on the same type of project (specialty) or if they migrate to another specialty.
Hope this helps :-)
So you have n subjects who provide the information whether they have stayed on the same project or switched to a different project. What do you need a test for? E.g., 300 members, 90 oft them switched, 210 did not. What do you want to test then? That (some) people switch is proven at once, if at least 1 sample member switched. You have no research question asked which requires a statistical test.
With kind regards
Karabiner
»Jetzt kann mich der Führer mal am Arsch lecken.« (Ernst Kuzorra, 1941)
So, I guess that the preliminary research question is: what proportion remains i the same position? So if there were 1000 first and 700 remain then the proportion p would be 0.70 and the confidence interval by:
p +/- 1.96*sqrt(p(1-p)/n)
A natural descriptive would be a table with the number of person in A, B, C etc as rows in the first period and the number of persons in the second period as columns.
(Of course the rows and columns would be statistically dependent so it would not be very meaningful to do a chi-squared test or Fishers exact test.)
Later on it could be interesting to look at transition probabilities of moving from e.g. A to B.
It could be interesting to look at if the risk of leaving A etc, is constant, increasing or decreasing. Like a sort of survival analysis.
Thanks a lot for your answers.
A survival analysis would, indeed, be interesting.
Tweet |