Money in chess

[Phil Thomas]

Hi again, you say

"Hi Walter. That's the reason for using regression analysis - determines causation and not correlation."

I don't think this is true Jonathan. Regression gives the best parameters should a correlation exist (and a measurement of the statistical strength of a correlation), but cause and effect are not included; they are always a matter of interpretation. That's always the most important part of any statistics! Cheers

Walter,
totally agree with your comment about interpretation - for any statistical work. Classic example in my mind is that one can look at the amount of damage done in a fire and correlate with the number of fire engines attending. If not careful you would conclude that the fire engines are causing damage.

However I strongly applaud the attempt at analysis by Jonathan. By definition he is not handling data that is not from a controlled experiment and there is every chance that key controlling parameters have not even been measured.

My work based experience with that kind of data is that you start off talking about the strongest correlations and therefore start off by telling the customer what he already knows.

In this specific case
Strongest result links high prize funds with large number of entries. Makes sense to me - in general prize funds come from entry fees. Tournament tend to budget from size of last year's entry which should be a good predictor of this year's entry.

Second strongest result - if I paraphrase it correctly - titled players are more likely to enter when field is large. This would be necessarily so if titled players are driven by the same factors as everybody else.

Jonathan,
what would be really interesting would be the next strongest findings from your analysis.
Especially interesting if they are non intuitive results.

Talking about controlling factors not listed.
Great playing room for the Edinburgh congress this year makes me more likely to play again next. For me it scores for view from window, acoustics, comfortable ambient temperature without draughts, comfortable chairs. Four things that can't go into any statistical modelling without incredibly detailed questionnaires - which most of us wouldn't bother to fill in anyway.