We recently conducted a research for one of our client's ad effectiveness campaign. We had asked questions to a control group who were not exposed to the ad and the same questions to the test group that was exposed to it.
Our standard practice it to collect at least 300 responses. We had N=321 for the Control and 387 for the test group. The concern was the difference in number of respondents between the two groups (387-321) = 66.
Does anyone know if this difference will skew the results too much? We are providing results at 95% confidence level.
Any insight is highly appreciated.
What method are you using to analyze the data? For t-test you can have unequal sample sizes. This was covered here in the past. http://www.talkstats.com/showthread....l-sample-sizes
From memory in certain cases of ANOVA not having the same subgroup n can complicate analysis of the data, but the real leason here is that your question depends on which method you are used to analyze the data with.
"Facts are stubborn things, but statistics are more pliable." Mark Twain
Posting Permissions
|
|
Reply With Quote