Sorry OP, this post is a little off-topic but I wanted to ensure good statistics are mentioned here.
WelshShooter,
If your velocity spread is 40 fps and change per 1/10th grain is 5 fps per Sierra's data, you'll need to shoot at least 20 shots (30 is much better) with each to have statistical confidence above 90% to show a 2/10ths grain change in charge weights changes average velocity 10 fps.
How many shots per charge weight did you shoot? 5 shots have about a 30% confidence level for anything shooting groups or measuring muzzle velocity.
Hi Bart, what program or table are you using to generate your sample size and confidence level? I'm using Minitab v16 to determine sample size.
For my testing I shot 5-shots per measurement group but upped my charge around 0.3gr per time. Minitab can perform a calculation for me which requires two of the three following inputs, along with the standard deviation:
1. Sample size
2. Difference
3. Power value (i.e. confidence interval)
It will then calculate the one field which is left empty. Since I had a sample size of 5, and I want a confidence interval of 95% (so a power level of 0.95) then I can detect an average velocity difference of 28.9fps. The standard deviation of the velocity was 13.1fps.
I completely agree with your thought on increasing sample size. If I use the above rationale:
1. Sample Size = 30
2. Difference = (we want to find out)
3. Confidence of 95%
We can then use this information to detect an average velocity change of 8.92fps with 95% confidence (based on the same standard deviation of 13.1fps).
Minitab (and myself) don't care about how much you want to change the charge weight. The 2-sample t-test is a statistical test which compares the mean values between two groups (e.g. two different charge weights, two different bullet brands with the same charge weight and so on). The 2-sample t-test produces results with a 95% confidence band, therefore the reasoning behind the 95% confidence for sample size as above.