Sid writes: > Well if the new "calculated" ranking goes into the model for > estimating an outcome, with other parameters remaining equal, and the > forecasts are 10% more accurate, that is 10% more results are > predicted correctly, then in the "real world" I would think it fair to > suggest that the ranking was 10% more accurate.
Your response is very vague. I'm not sure why this is so, but I can guess: 1. You are trying to use an approach that will work for many different situations, so you haven't yet come up with a common approach that applies to all situations. 2. You've just started working on the problem and haven't yet had time to provide firm operational definitions. 3. You are hoping to patent your approach and need to keep your competitors from finding out what you're working on. 4. You are trying to understand the problem from a very abstract level. 5. You are involved in some sort of illegal sports betting. The bottom line is that the quality of advice that you receive is directly proportional to the amount of information you are willing to share. Of course, with the limitations of email, even under the best of circumstances, your advice is likely to be mediocre. But when you can't provide any specifics, the quality of advice that people like me can provide is often worthless, other than providing us both with a bit of entertainment. You want to use "the model" to estimate "an outcome" with "other parameters" held constant. What type of model? Is it a regression model or a time series model? What type of outcome? Is it categorical or continuous? How much data do you have where this outcome is known? Is there more than one outcome? What are the other parameters? > There must be a valid way of ranking one team over another or others? You might want to define "valid". And you tell us whether your primary goal is ranking, prediction, or validation. These three very different goals would dictate three very different approaches. It might be a combination, of course. By the way, there is more than one valid way to rank a team just like there is more than one valid way to assess whether a child has a fever. And there is no "best" way to rank a team, just like there is no "best" way to measure a fever. I hope you don't interpret all these questions as an attack. It is very hard to define these terms well. Quite often, the best thing that a statistical consultant can do is to get the client to characterize his/her problem more precisely. Steve Simon, [EMAIL PROTECTED], Standard Disclaimer. The STATS web page has moved to http://www.childrens-mercy.org/stats. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
