we want to compare the sensory acceptability of native vs. broiler chicken meat in terms of taste, texture, etc. samples of meat were only acquired from the local market, i.e., this is not a controlled experiment. what reservations as to the conclusions should be considered?
You are comparing two types of chicken meat. If you find a difference in acceptability between the two types this can be due to many reasons like age, nutrition, slaughter procedure etc etc and due to genetic differences between native chickens and purpose bred broilers. But do you need a test to prove that?
Thanks for your response Kees. Basically, we want to determine if there any significant difference on the acceptability between native and broiler. however, I am not sure if its appropriate to do so, given that the study design is not experimental.
Well, You will eventually have a significance, that will just tell you whether there was a difference or not in whatever parameters between the two batches you were comparing. But this will not allow you by any means to make extensive these results to the population of native chickens or broilers
Does every participant get both native chicken and broiler? Then paired t-test or w-test are the right choices (depending on the scaling of the data (interval vs ordinal)).
I think that you can generalize your significant (and not significant when there is enough power) results to the population (this is not the problem, because there is no problem with your statistical inference test). You can say, that these two sorts of chicken differ in your DVs (you could not do that if the two sorts of chicken were not randomized to the participants, but I think your problem is that you cannot randomly assign a chicken to the group native or to the group broiler). Because of the quasi-experimental nature of your study, you have problems with your internal validity, so it is not really clear why the two groups (native vs broiler) differ in their taste etc. There you should try to collect as many reasonable control variables as possible.
Dear Christian
Your chose depends on your data , you can use T- test with scale and normally distributed data while in a case of ordinal or not normal distributed data you can use non-parametric test , in this case, preferred Mann-Whitney test because you have independent samples while paired T- test and Wilcoxon signed-rank used for dependent samples.
Good Luck
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis