We have two nominal variables: diet and location. We want to know whether the diet of the species differs in various locations. I know it is possible to do it by chi-square, but I also have seen others have done it through the Kruskal-Wallis test. I applied chi-square, though I still believe Kruskal-Wallis is the right choice. Am I right? However, I have difficulties with performing this test on my data. Here is a view of a little part of my data.
Maybe I should re-order my data in another way?
I really appreciate it if you could help me.
If the data really are nominal then a chisq could be appropriate. But, what is you location variable? If these are locations that, for some, will be next to each other, then it is likely diets will be more similar in those than if more distant. Or is location just something like desert, mountains, ocean?
If the data really are nominal then a chisq could be appropriate. But, what is you location variable? If these are locations that, for some, will be next to each other, then it is likely diets will be more similar in those than if more distant. Or is location just something like desert, mountains, ocean?
Adding to Daniel Wright's answer, the Kruskal-Wallis test could be used if your explanatory variable was nominal and your outcome was something that can be ranked (preferably without too many ties). If the outcome variable is also nominal (as you say yours is), it cannot be ranked. Therefore, the KW test cannot (properly) be used.
HTH.
Dear Shabnam Shadloo
Interesting question. Waiting for more answers from experts.
It is not proper to use Kruskal Wallis in your study. But if you want to analyze your data using an additional test, you can use Cramer V correlation analysis. Because it shows not only the chi-square result but also say the level of your correlation result.
Good luck with your study.
Dear Daniel Wright , thanks for your answer.
The locations are two rivers flowing towards a sea, a delta, and a beach. I do not believe they are dependant, especially the two separate rivers, are they? and what can I do if they are?
Okay, so your n=2 for the two locations. What are you units? Are they independent?
Hi Shabnam.
It looks to me that chi-square is the most appropriate test for your data: you are looking for an association between two nominal variables (i.e. diet and location). Because of the way you've entered your data you need to weight your cases by your "count" variable before running the analysis - either via the GUI (Data > Weight cases...) or by entering the following line in your syntax:
WEIGHT BY count.
before your CROSSTABS commands.
Best of luck!
Roxy
Daniel Wright actually n=4, and I didn't understand your question. Which variable's unit are you asking about?
Shabnam Shadloo , so you are comparing two locations. The question was what are you observing. For example, if I am comparing two universities, and have a hundred students sampled in each, my focus could be on students or on the universities. Say I am interested in much each students likes physics classes. Well, the students at 1 university may like it because one of the teachers is great, but these students then are not independent. You'd need to consider also teachers. The commentators who are saying chisq is alright, are assuming much about your design. You might think about starting a new question where you provide more information.
Shabnam, without running any test, I could pretty much tell from your numbers that your result was not 'significant'. I use the term here not in the null hypothesis testing sense, but simply to convey that the outcome of your experiment a) can be a matter of chance, b) may not inform the question you were asking to begin with (i.e., not a statistical question), c) you did not supply enough information, as Daniel Wright says.
I went ahead and made a contingency table with your data. I tried to come up with two nominal groupings of diet and location and came up with the examples in the spreadsheet. Let's imagine that, by looking at receipts or by some other retrospective measure, you counted where and how many times you ate the above foods over a month. Whether this is an appropriate design for a test sample is beside the point - you can substitute what is counted with something else - the test doesn't care and the outcome is identical. The interpretation should be predicated on the goals of the study.
I might have misinterpreted your data, though, but perhaps this example will still be useful.
Anyway, what's to be made of the results: you don't seem to like mashed potatos very much, so ate them little, regardless of location.
You like eating restaurant tacos and pizza a lot more than homemade (or take-out, whatever). The differences are pretty large, but you still eat at home quite a bit. Statistically speaking, as evident from chi-square, this is all within the realm of chance. Of course, since we framed this for an individual, we can disagree with the test. If your experiment relies solely on Fisher's set up of randomness, then your experimental results are null. You are free to explore other means of scientific description.
And to put my question into Miky Timothy 's example. Suppose it is Clayton's birthday and he is having 20 other friends to show up. His mom is deciding whether to have the pizza and tacos at their house, or at a restaurant, and it only depends if it is raining (food would be in the backyard). It rains, so those 21 go to a restaurant. How does this affect your conclusion?
ps. Obviously the CLayton example is extreme, but anything where knowing that Clayton went for pizza affects your probability of knowing Mookie does would also affect the conclusion. The data would not be independent (and this is an assumption of chisq).
There are 266 sample of otter feces which we collected throughout a year, and we have analyzed the species diets through their feces. I think they are independent. I cannot think of another factor that can affect them, besides their locations.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS