Originally posted by coon But these 5% are equivalent to about 45 points (with a mean of 899 points). And the difference between Pentax and the industry standard is only 11 points or 1.24%. So not very meaningful.

I also did not calculate anything "correctly", I just applied the estimate 'uncertainty = mean/sqrt(N)', and that gives about 32 for mean=899 and N=810 (number of samples). It was more or less a plausibility check only.

coon

Ah. Wrong equation. There are two different questions that need to be answered:

1. Is the Pentax sample size sufficient to make predictions about the entire population? and

2. Is there sufficient information to determine if the rankings between Pentax, Nikon and Canon are "statistically significant" i.e. that you can say one population of customers are in fact happier with their cameras than another based on the data collected.

To answer question 1, you determine the

confidence interval based largely on the sample size. If based on your earlier assumption of 810 pentaxians AND an assumption that the sampling is random, the sample mean would be within 3.4% of the population mean 19 times out of 20. Once you get above a sample size of 200, the calculation is fairly insensitive, and you can get to within 5% of population mean regardless of the size of the population.

To answer question 2...talk to a stats prof, all I know is it requires some kind of test called ANOVA and requires more information than we have from the study. Although I think that in this case, yes, it would be unlikely that the differences between Pentax, Nikon and Canon are statistically significant, or at least that I wouldn't assume that they were until someone showed me the numbers.