Hi,
I've got a dataset compounded by 307 observations, i.e. answers made by individuals which were asked whether or not they take an action . Reponses are binary therefore (Yes I do / No I don't). 5 actions were asked.
I've implemented a Cochran's Q test in order to test the difference between the 5 proportions. I've implemented it in R Studio with 2 different packages (NONPAR and RSTATIX). Both of the r functions lead to the same result with a very high test value (Q = 450, df=4) indeed , conforting that there is a significant statistical difference between the 5 proportions. Note that such result is consistent with the descriptive statistics (proportions range from e.g 20% for the lowest to 85% for the highest). Nevertheless this is the first time I got such a high value for a statistical test
Did I missed anything ?
Thanks for your help.
I assume you mean that you have 307 individuals, so in long format you would have 307 x 5 = 1535 observations. If so, it appears that a perfect dichotomy in responses across actions would yield a chi-square value of greater than 1000.
"Q" and "Chi-square" are the same for this test.
An example in R follows.
# # #
Practice = factor(rep(1:5, 1, each=307))
Respondent = factor(rep(1:307, 5))
Response = factor(c(rep("Yes", 307*2), rep("No", 307*3)))
Data = data.frame(Practice, Respondent, Response)
library(coin)
symmetry_test(Response ~ Practice | Respondent,
data = Data,
teststat = "quad")
library(RVAideMemoire)
cochran.qtest(Response ~ Practice | Respondent,
data = Data)
library(rstatix)
cochran_qtest(Response ~ Practice | Respondent,
data = Data)
Than you very much for your answer Salvatore. If I got it right it confirms I did not bad ;-)
best regards
/Jeoffrey
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude