Hello! I would like to address the experts regarding a question about conducting a statistical analysis using only nominal variables. Specifically, I would like to compare the responses of survey participants answered the question whether they take certain medications "Yes" or "No", and analyze the data with different criteria such as education level, economic status, marital status, etc. I have conducted a Chi-squared test to determine if there is a significant difference between the variables, but now I would like to compare the answers of whether or not this medicine is taken depending on each group, for example in the education variable (higher, secondary, vocational and basic education). Is there a statistical test similar to Tukey's test that is suitable for nominal variables? I would also like to know if it is possible to create a column chart with asterisks above the columns indicating the significant differences between them based on this test for nominal variables.
I usually use Statistica StatSoft and R studio. But none of my attempts to do post-hoc for nominal variables analysis on any of them were successful. In R studio I tried pairwise.prop.test(cont_table, p.adjust.method = "bonferroni")
But I got an error:
Error in pairwise.prop.test(cont_table, p.adjust.method = "bonferroni") :
'x' must have 2 columns
I assume that this is due to the fact that I have groups in one of the variables and not two.
What should I do?
Thank you in advance for your help!
Hi,
The following videos will clarify your doubts:
https://www.youtube.com/watch?v=cOu9rv83G-I
https://www.youtube.com/watch?v=krLz0GK3uwg
https://www.youtube.com/watch?v=Rp0qorrPXA0
After performing a chi-square test on nominal variables, you can use post-hoc tests to determine which categories have significant differences. However, unlike ANOVA, no commonly accepted post-hoc test for chi-square analysis exists.
One option is to use pairwise comparisons of the categories using a series of two-by-two chi-square tests. This involves comparing each category with every other category to determine which pairs are significantly different. However, this approach can lead to an increase in the familywise error rate due to multiple comparisons.
Another option is to use adjusted residuals or standardized residuals to identify which categories have the largest differences. Residuals represent the differences between the observed and expected frequencies under the null hypothesis of independence. Large residuals indicate a greater deviation from independence and suggest that the category is contributing to the overall significant chi-square test. Adjusted and standardized residuals can be calculated in most statistical software packages and used to identify which categories have significant differences.
In addition, graphical representations such as bar charts or mosaic plots can also be helpful in identifying which categories have significant differences.
It is important to note that the choice of post-hoc test may depend on the specific research question and the nature of the nominal variables being analyzed. It is recommended to consult with a statistician or a subject matter expert to determine the most appropriate method for analyzing your data.
I did not check the R function indicated but the error message seems to suggest that it may have been expecting a data frame where something else was fed it. The rest I do no find very very clear. I an not even convinced you need multiplicity adjusted pairwise comparisons. However you report that you are in a medical university, a sort of entity which should have a biostatistics unit. If so, I suspect it will be more productive to set up a meeting with that unit, starting with a careful description of the study, before moving to statistical questions.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS