Hi all, I am not sure what data analysis will be suitable in the following study, would be grateful if someone could provide suggestions for that:
The survey is used to figure out what fruit shall be promoted in the next marketing campaign. We try to understand the preference on fruit in general, as well as the preference within each target group. We provided 5 options and ask the customers to rank the options from 1 – 5, with 1 being their 1st preference. My questions would be, what data analysis would be suitable in the above study, as I am struggling between the crosstab chi square test and the friedman test. As for the variable value it is ordinal (1st, 2nd, 3rd) but then the variable itself (eg banana, apple) is nominal.
Target groups: Kid | Teenager | Adult
Options: Banana | apple | orange | Pear | berry
Rank: 1,2,3,4,5
Many thanks!
This question was on researchgate:
https://www.researchgate.net/post/How_to_perform_data_analysis_for_ranking_of_familiarity_of_foods
The Friedman test should yield the same results as the Chi-square test, I think. But, indeed, the Friedman test accounts for the rank order.
You can use the (nonparametric) aligned ranks Friedman test. In R, with the scmamp package. See example in attachment. https://github.com/b0rxa/scmamp/blob/master/R/tests.R https://cran.r-project.org/web/packages/scmamp/scmamp.pdf In SAS, see the 2nd example in attachment. By the way, your data are ipsative, i.e. the sum of the rankings, for each participant, is constant. If there are, e.g. 5 foods to be ranked, the sum is 1+2+3+4+5. Greer & Dunlap have argued one can run ANOVAs on such measures. Citation: Greer, T., & Dunlap, W. P. (1997). Analysis of variance with ipsative measures. Psychological Methods, 2(2), 200-207 .
Jos Feys Thank you the post! I mainly use SPSS for data analysis but I will definitely explore R. Given some of the ranking data is incompleted, for instance, participants does not rank their preference from 1 to 5, my current approach in SPSS will be replace it with 0 since that may also be one of their responce. Is that approach also valid in R or do I have to change it to NA? thank you
I can help with the non-response part. I don't recommend changing non-responses to 0 because then those responses will be mean that the rater has provided a really low rating.
Instead, see this link:
https://www.ibm.com/support/knowledgecenter/SSLVMB_24.0.0/spss/base/define_missing_howto.html
- Badges
- Science topic