For example, I have a data set of drainage classes with four levels (e.g., a, b, c, d). It was measured from randomly selected points in two different areas (e.g., East and West). What I want to test is whether drainage class in these two areas are different.
At first, I tried the Chi-square test for independence with two variables: drainage class (4 levels) and areas (2 levels). However, I think that the test statistic calculated from the test just shows if these two variables are associated, which is not what I want to see. What type of statistic analysis can I apply to my case?
Hello Jeongho Han. From what you've said, a Chi-square test of association sounds okay to me (assuming there is no inherent order to the 4 drainage classes). If your concern is that the Chi-square test does not distinguish between explanatory and outcome variables, I don't think that matters particularly.
But if you really want to use a model that does distinguish between explanatory & outcome variables, you could estimate a multinomial logistic regression model (assuming no inherent order to the 4 drainage classes) or an ordinal logistic regression model (if the 4 drainage classes do have some inherent order). HTH.
Dear Jeongho
Bruce is right - the Chi-square test would work. If I were you and if data on drainage is scalable (not categorical) data, I would try the ANOVA test - use regions as your IVs and drainage data as DVs.
Best
Eugene (Yevgen)
In my opinion if you have three variables
and also
randomly selected
So, think of using a mixed model.
Thank you everyone for all answers but I'm a bit confused in interpreting the result.
For example, if I performed a Chi square test and the resulting p value was < .001, I can conclude that H0 is rejected, meaning that the two variable (drainage class and areas) have a significant relationship. If my interpretation is correct, can I interpret this result in a different way that drainage class depends on regions (i.e., East and West) and thus West and East have different drainage class?
Jeongho Han Look at the attached pictures. It is on the same data as you.
Jeongho Han , Yes, and this isn't always intuitive the way we talk about the results of this test. Saying that there is an association between Area and Drainage Class also implies that there is a difference in distribution of Drainage Class between Areas.
I'm uncomfortable with you saying that there is a "different drainage class" between Areas. Instead, what you've found is that there is a difference in the distribution of counts for Drainage Classes between Areas.
Because of this, be cautious what conclusions you draw. As mentioned in earlier comments, is Drainage Classes actually an ordered variable ? Is the difference in distribution in Drainage Classes meaningful ? I added a plot, that, if I made it correctly, suggests that the big difference is between B and D across Areas.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA