Hi All,
We are considering a CBC for the small population (pediatric cardiothoracic surgeons) there are about 100-120 of them in the country we want to test. Maximum sample we can get is around 20. Study contains two attributes: Product (7 levels) and Price (6 levels). We think about 15 tasks (4 cards per task + none) . Constant sum allocation.
In theory for statistical significance (5% margin of error) we would need to test 92 respondents. It's unrealistic. We agree that the data will not be statistically significant but what worries me more is the standard error for the main effect is also high. Does it mean we have to forget about CBC for small populations? Even if we test 1/5 of the population? What if a population is 50 and we test half of it? Doesn't HB help?
I would be grateful for your opinion.
Robert