# How to test significance of frequency in one expression of a rating scale?

Hello,
i think i have a problem that is rly easy to solve, maybe i just got a bad day or sth..
Ive made an experiment with 4 groups, below you can see the answers of a rating scale 1 to 7.
The value -1 is just a control answer, that should mostly occure in the control group (= KG)
Obviously it does, but how to check significance?

``````> tally(WN~group, data = data_reduziert)
group
WN   EG1 EG2 EG3 KG
-1  16  21  29 41
1    6   3   0  0
2   12   6   4  1
3   11  14   3  3
4   31  35  24 21
5    9   3   9 10
6    2   5   6  7
7    1   3  16  7

Ive tried chi square test:
> chisq.test(table(data_reduziert\$group, data_reduziert\$WN == "-1"))

Pearson's Chi-squared test

data:  table(data_reduziert\$group, data_reduziert\$WN == "-1")
X-squared = 18.341, df = 3, p-value = 0.000374

is this the right way? But i also want to know between which groups it differs: so i need a Post hoc test
How do I do that? chi.posthoc.test didnt work
``````

I would suggest consulting a local statistician or one of your departmental methodologists. You are asking a statistics question not an R question. You really need to sit down and discuss program design. In some ways checking for significance is the last thing to worry about.

Without knowing a lot more about the study it would be irresponsible to suggest anything specific but typically (traditionally?) one treats a 7-point scale as continuous rather than categorical. So a chi-square test seems rather inappropriate.

Sorry I cannot be of more help.

atm is semester break, so it´s hard for me to get some advice there xD

thing is that -1 is to understand as "neither nor", so i only transformed it as a categorial variable to proof that in the control group there was no stimulus to rate from 1 to 7.
After i proof this, i cut out the value -1 (mutate to NA) and change the scale niveau back to numeric to make an ANOVA to proof the participants perceived the ad stimuli in the desired manner, thats then no problem.

I thought about to make a binomial test, therfore i would dummy code the values so that -1 = 1 and else = 0. Can u tell me how to deal with the if.else function?

``````WNTest %>%
mutate(WN = recode(.\$WN,
"-1" = "1", ifelse(.WN, = 0)")) -> WNTest

``````

I´m pretty much of a beginner in R

thing is that -1 is to understand as "neither nor", so i only transformed it as a categorial variable to proof that in the control group there was no stimulus to rate from 1 to 7.

Statements like this are even more confusing!

Do you have a draft of your intro &/or methods section? If you stick it up on a server such as dropbox or mediafire and post a link here someone might have a look at it. And maybe include a couple of references that might give us some idea of tee subject area. At the moment me are in an information vacuum about what you are trying to do. For all I can tell maybe you should be considering a factor analysis or a Bayesian approach.

It also would be god if you could provide your raw data or a sample subset of it if it is really big) in dput() format. I would not try any analysis until you have some concrete suggestions on what might be appropriate. It it all too easy to plug some numbers into a function and get total crap out (but which looks nice).

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.