Originally posted by JeffHamm
View Post
Thank you for all your hard work. I would beg your indulgence to make a few comments. The sample size for your model is, as you have said, very small and cannot be considered to be statistically significant. I haven't before seen a study with a component sample size of 2. When I did my Surveying Degree I wrote a computer program to compute astronomical observations. I was unable to declare it successful until it had been rigorously tested against known results. We have available a small sample of 5 (or7) PMI's with known results, where projected margins of error obtained from your model can be compared the achieved margins of error. There is 1 sample for Nicholls, 2 for Stride, 2 for Eddows. It would also be interesting to compute the error margin for the PMI for Kelly to see if Maxwell can be dismissed on the basis of Bond's estimate. I read somewhere that Phillips also did a PMI for Kelly but I haven't been able to find a reference.
Best regards, George
Comment