Anemone wrote:
If people want to check, you can calculate the t-values yourselves. I was using [(control means - WP means)/control SD] (I used Rogers et al SDs for both Rogers et al means and Davis's means, since he didn't post SDs. If you're supposed to use a different formula, go ahead and crank it out and let us know. I'm going by memory as to how to do it. I mean, I think the trend is pretty obvious, so I'm not too worried.
OK, I'm just checking this as I go along and am not overly familiar with it, so might make a mistake. I'm testing the means for the WP diagnosed male group (n1=20) and the combined Rogers control group (n2=21, 18% female),
for the sample sizes and variances unequal (hence the difference in the denominator: standard error of difference between means). This is for perspective taking.
I chose level of significance to be 0.05.
T = (18.9 - 11.70)/sqrt[(4.3^2/n2) + (7.10^2/n1)]
= 18.9 - 11.70 /1.84
= 3.91
The number of degrees of freedom is over 30, so using a t-test table, our T value is greater than the table value (of about 2.03) at sig. level 0.05. Therefore, the means are statistically significantly different at that level.
I quickly plugged the means and SDs into a t-test calculator, and it gave a similar result, with T = 3.95 at the 95% confidence level. I could check these with a few stats routines, but don't have time.