| |
BGonline.org Forums
-1
Posted By: Casper van der Tak In Response To: Timothy (Bill Phipps)
Date: Friday, 29 June 2012, at 7:38 a.m.
Bill, you don't know Timothy, your accusation is wrong. For instance, in the field of BG analysis, Timothy introduced the concept of formulating hypotheses why one play is right, and then changing the position slightly to test the position - what some have called 'Chowing' the position. I think your accusation is the opposite of what Timothy is about.
I think Timothy is worried about wrong inferences not from you (Bill Phipps), but you (universal, the world at large). He is also right that even intelligent people will draw the wrong inferences from statistical data and reports, and that caveats are removed all too often.
I myself have seen social scientists acquiring a data set, run a series of regressions analyses with different variables and different functional forms, and then report a 'statistically significant' result as if the finally selected regression had been the only one ran. Of course, this leads to spurious results. I suppose that this type of data-mining can only be used to formulate hypotheses that can then be used in new statistical tests with new data sets.
So you can formulate your hypothesis on the basis of the data you have: "Sticks PR results can be fully explained assuming his results are drawn from the same PR result frequency distribution as ... (name the peers)" You then gather data, specify the frequency distribution by analyzing the PRs of the peer group, and then test whether the hypothesis re. Stick can be rejected at a pre-chosen confidentiality interval.
This, I think, would be a safe analytical approach. I cannot really think of another safe approach.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.