How does having unequal sample sizes effect my statistical conclusions?

Question

I have 2 groups, one who received treatment and a control group. About 12 variables were measured in each group. The Treatment group has about 65 subjects, the control group only has 25-30. Is it better (statisticaly speaking) to have the groups as is (65:30) or should I take a random sub-sample of the Treatment group and make it even (30:30)? Seems to me like more=better, but I'm not sure if that is true or not.....

so, how does that effect the power?

Answer 1

I would stick with the "more is better" idea. Why throw out data? If you do decide to eliminate (good) data, then you need to make sure those you retain are random, like number them and then use a radnom number selector. In other words, don't throw away 35 pieces of data just so that the 30 that remain are the best in the treatment group. Need to make sure it is a simple random sample (SRS).

More is better

Answer 2

The larger your statistical sample the smaller your confidence interval would be, and therefore the more accurate your analysis. it would also depend on what you are trying to analyse and how quantifiable the result is. Given that you have such a high number of variables you are measuring, you are better off keeping the groups as they are. Otherwise, by reducing your treatment group, you may be inadvertently biasing your selection and skewing your results.

Answer 3

i imagine that larger pattern sizes have a tendency to develop the chance that the predicted recommend is on the point of the extremely recommend. yet i don't think of pattern length has an instantaneous result on statistical magnitude (that's determined for the try) hence that's basically as person-friendly to get a statistically important outcome at the same time as N=22, as at the same time as N=1022.