I ran an experiment to evaluate fly attraction to a number of bacterial cultures (n=50). Briefly what I did is: 20 male flies were exposed to two odour sources at a time-one containing a bacterial culture and other containing control; number of flies that showed attraction to the culture and control were counted after a certain time. The test was repeated 10 times for one bacterium. Similarly female flies attraction was also tested for the same bacterium. Thus for one bacterium I had 20 replicates in total (10 for males and 10 for females). Now, I want to analyse the data to evaluate which bacterium showed attraction at what extent and how they differ gender-wise. I am not certainly sure which statistical test will be suitable for this purpose. I was thinking about Kruskal-Wallis test but not sure if its the correct one. I am using SPSS. Could you please suggest which statistical test will be suitable to analyze this data set?
Thanks in advance.
You could use the chi-square test. For one (each?) bacterium, the test was repeated 10 times; so, you could count the total number of flies who were attracted and not attracted, and this separately for male and female flies.
This could, for example, yield this crosstable (with the total counts in the cells):
----------Male----------Female
Yes--- 80--------------60
No-----20--------------40
In this example the chi-square is significant:
X-squared = 8.5952, df = 1, p-value = 0.00337
In this example the male flies were significantly more attracted than the females flies. With these residuals:
residuals male female yes 1.195 -1.195 no -1.826 1.826
See also the "jpeg" for these residuals.
I added the R script. You can run R code online:
https://rdrr.io/snippets/
For the chi-square using SPSS:
https://www.youtube.com/watch?v=wfIfEWMJY3s
Dear Nazma Akter Tithi and colleagues,
Based on your question I understood that you have a factorial design.
If this it´s correct, I recommend that you use a factorial ANOVA, in where you may have the following function: numbers of fly context selections in function of sex*type of context (bacterial vs control). You will have a 2 x 2 factorial ANOVA.
I recommend too that you use a robust version of this statistical approach.
In Patrick Mair and Rand R Wilcox paper, you can find a very useful robust option, essay to apply in R:
Article Robust statistical methods in R using the WRS2 package
Best wishes.
You can add the 'type of context' in the chi-square test with the Cochran-Mantel-Heaszel statistic in SPSS:
https://www.youtube.com/watch?v=ilKZG4nmuLU
I added this test in the R script for this example:
Context----------Gender----------Attraction----------Count
bacterial---------male--------------yes--------------------80
bacterial---------male--------------no---------------------20
bacterial---------female-----------yes--------------------60
bacterial---------female-----------no---------------------40
control------------male--------------yes--------------------50
control------------male--------------no----------------------30
control------------female-----------yes--------------------45
control------------female------------no---------------------35
The CMH-test is sign. here, which indicated that the Context yields different effects. For the bacterial context there is a Gender effect on Attraction, but not in the control context.
Mantel-Haenszel X-squared = 7.1069, df = 1, p-value = 0.007679
Dear Dr Jos Feys , Thank you so much! You showed me the right direction. I analyzed my dataset accordingly and it seems perfect :) I really appreciate your effort in helping me to solve the problem.
Dear Dr Pablo Ezequiel Flores Kanter , I appreciate your suggestion, however I have found chi-square test a bit more related and easy to understand. Thank you :)
- Badges
- Science topic