Can you explain what you're trying to do? Are you looking to do a t-test eventually? It's not very clear...
I have two samples and I was wondering if there is a way to standardize the calculations in my smaller sample so that the calculations in both samples would have the same weight.
Can you explain what you're trying to do? Are you looking to do a t-test eventually? It's not very clear...
The basic info is that I have Sample A (n=33) and Sample B (n=18) and I counted various occurrences of different things per individual in each sample (example: # of cavities per person). The two samples are completely independent of one another. Originally I calculated mean, median, and mode for all of the ailments, as well as independent t-test to check for significant mean difference between the two samples.
The issue I am now facing is that my professor wants to me to account for the fact that sample B is horribly small and figure out a way to compare the two samples in a more even (weighted?) manner (which probably means all of the calculations I already did are junk).
Here is an example of my data:
Sample A # of cavities/individual (n=33): 1, 0, 10, 1, 8, 3, 13, 5, 0, 0, 5, 0, 8, 7, 0, 2, 10, 7, 0, 0, 7, 0, 1, 1, 8, 2, 0, 2, 2, 2, 5, 0, 4
Sample B # of cavities/individual (n=18): 4, 4, 0, 1, 8, 1, 1, 0 , 9, 4, 5, 4, 1, 2, 2, 3, 1, 0
The overall outcome that I would like to eventually report is something along the lines of "yes there is a statistical difference between the two samples" or "no there isnt"
What's the difference between Sample A and Sample B? How does a person end up in Sample B?
Does your professor want you to consider matching a person in Sample A to a person in Sample B? eg a 30 year male in Sample A to a 28 year male in Sample B. Maybe you have some other data on the individuals in the sample, like time since last dentist visit.
Both samples are historic skeletal. Sample A is Dutch colonials and Sample B is Native American. Both Samples are taken from the same geographic region, which is the reason why I'm comparing them. The aim is to see if the dietary or lifestyle differences between the two group affected their dental health.
For my research, I looked at the two sample collections and did basic counts of the different ailments as I saw them in the skeletons...so I don't have other data besides what I saw.
Hi!
In my opinion, it not impossible to work with small samples inasmuch one bears in mind what this implies (i.e. less power; see: http://en.wikipedia.org/wiki/Sample_size; http://www.jeremymiles.co.uk/misc/power/).
As for your samples' selection, it recall me the oft-quoted issue of sampling (and hypothesis testings) in archaeology. There is a good literature on this topic. I would suggest to give a look at the book of Drennan (http://www.amazon.com/Statistics-Arc...dp_ob_title_bk).
Best Regards
Gm
Your sample sizes are not extremely small, and the unequal sample sizes not a serious problem. There is no "weighting" to be done that is not taken care of by the standard tests. However, your count data are not normal (maybe Poisson), so a t test is not technically the appropriate test. You might consider either a non-parametric test (Mann-Whitney) or a specialized test for comparing two Poisson means.
Tweet |