Goodnight. I am making comparisons of proportions of households that consume different vegetables. I am comparing the proportions of consumer households of each vegetable corresponding to each decile of the national distribution of household spending. I'm doing them with Z tests with pairs of proportions. I am not aware of multiple range tests to compare proportions similar to those used for multiple comparisons of means, such as Tukey or Duncan. Does anyone of you know of a test that might help?
It would be helpful if you could provide a small toy example. ... Is it the case that an omnibus chi-square test of association would be appropriate ? ... It may be the case that running multiple z-tests of proportion, perhaps with a p-value correction for multiple tests, would be an appropriate approach.
Dr. Mangiafico, thank you very much for the hint. Last night I found a test that in Spanish is called the "Marascuillo procedure", and it is precisely a chi-square test that indicates whether there are significant differences in the group of proportions. Then, comparisons of means are made with Z tests corrected with chi square.
SORRY, on the last line it says: “Then, comparisons of means are made with Z tests corrected with chi square”, and should say: “Then, comparisons of PROPORTIONS are made with Z tests corrected with chi square”.
You might see the following. I'm not sure I entirely understand the methods they are comparing. But it sounds like what they did was use component 2 x c tables from a larger table, and then adjust the results with a p-value correction methods. When compared with the Marascuillo procedure, Marascuillo was excessively conservative compared with a couple of these methods. But there is a note at the end about not using the p-value correction method when c > 10. (Although I don't entirely understand their point for this note.)
Article Comparing Performances of Multiple Comparison Methods in Com...
I did an exercise with Z tests for two-by-two comparisons and with the Marascuilo method. The results are similar. For familiarity with Z tests, I'll end the results with these. Attached the spreadsheet. I hope you read a little Spanish.
Buenas tardes. I was working on the layout of a book that I want to publish and had moved away from comparisons of proportions. You showed interest in their data. I am sharing a file with information on the households that bought green peas. The source of the data is the "Encuesta de Condiciones de Vida de 2014", Guatemala, Instituto Nacional de Estadística.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression