Statistic test in JMP

Question

I assigned a color to 1000 samples of soil from 3 different location. Now I need to compare the color at 3 locations to see if there is statistically different or not? The only data I have is the count of each color and from which location, my research question is that if the location affects on the color?

for example

gray color: 80 at site1 - 50 site2 - 75 site3

Brown color: 12 at site1 - 5 site2 - 32 site3

and so on....

what statistical test I need to compare?

Thanks

Mark_Bailey · Answer

OK, then Color is in the Y role and Site is in the X role. Use Count in the Freq role.

MarginalPanther · Answer

Thank you very much for your answer. I did the same test but I choose site and color both as a x factor and I put count as a response which I thought doesn't make sound.

I collected the samples from the river bed and it was from upstream- Midstream and lowstream, the color assigned to each particles by comparing with the Muncel Chart. The percentages of the color of particles are different from each location whcih might be because of different origin. I want to know if there is a different in color statistically and if there is,  which location or which color are responsible for this different?

Mark_Bailey · Answer

I will answer your question after making a lot of assumptions. You should verify these assumptions before trusting the analysis that I suggest. See the questions at the end. First, let's set up the analysis.

Here is the JMP data table that supports a contingency table analysis:

Select Analyze > Fit Y by X and assign the data columns to the specific analysis roles as shown here:

(I might have mistaken the analysis roles. Perhaps Color is the response and Site is the factor.)

Click OK to launch the Contingency platform:

This analysis compares the counts of the colors at the sites to what would be expected if there were no association between Site and Color.. The whole model test is based on a chi-square distribution of the test statistic. The test statistic is obtained using two different methods: Likelihood Ratio Test (LRT) and Pearson chi-square.

The assumptions have to do with how the data are collected. What do you mean by "I assigned a color to 1000 samples of soil?" How were the three locations selected? What is the purpose of the study? How is color implicated? What are you counting about each color and site combination?

To be honest, you should reverse engineer your study. Start with the desired outcome and work backwards. What is the research question? What analysis will provide the answer? What data will best support the analysis? What are the factors and responses in the scope of this study? it is unwise to collect data first, and then ask how to analyze it. Your data might be worthless.