# which statistical test should i use for my project?

i have conducted an experiment with two conditions (happy and sad) i wanted to see the effects of these conditions on words chosen from three different categories (positive, negative and neutral words)

-in both conditions i have data on the three different word categories

its a between subjects design - their were different paricipants in each condition the IV is the mood condition either happy or sad

the DV is realtive frequency of words in pos, neg and neutral categories

i want to find a statistical test which will show me if there is a difference between the realtive frequencies in each word category in the two different conditions

Maybe make a table with rows Happy/Sad and columns Pos/Neg/Neut. Find word counts for each of the $$2\times 3 = 6$$ cells.

Then do a chi-squared test to see if Pos/Neg/Neut choices are homogeneously distributed for Happy and Sad populations.

To illustrate how chisq.test works in R, consider the following table with fictitious counts:

TBL = rbind(c(21,34, 53), c(65, 27, 10))
TBL
[,1] [,2] [,3]
[1,]   21   34   53
[2,]   65   27   10

chisq.test(TBL)

Pearson's Chi-squared test

data:  TBL
X-squared = 52.536, df = 2, p-value = 3.909e-12


For these data the null hypothesis of homogeneity is strongly rejected with a P-value near $$0.$$ So it seems Happy and Sad people are picking different kinds of words.

• If you get an error message for your real data complaining about inability to get a reliable P-value on account of low counts in some cells, you might be able to use the argument sim=T in chisq.test to get a useful P-value.
• If you wonder whether additional tests might be useful to try to track down what specific differences in word type there are between Happy and Sad people, you may get some ideas from similar sources. Maybe use $ notation chisq.test(TBL)$resi for clues from 'Pearson residuals'.