Changes

:: OK, we agree there won't be a double-blind study, or any formal study.  But my question remains unanswered:  what increase in cancer incidence would you require to admit cause for concern?  100x?  1,000,000x?--[[User:Aschlafly|Aschlafly]] 23:01, 19 May 2008 (EDT)

:: OK, we agree there won't be a double-blind study, or any formal study.  But my question remains unanswered:  what increase in cancer incidence would you require to admit cause for concern?  100x?  1,000,000x?--[[User:Aschlafly|Aschlafly]] 23:01, 19 May 2008 (EDT)

Well, you didn't ask me the question, so I didn't realize you wanted me to answer it, but ok.  There are two things that I would want to see.  First is whether the effect of Hollywood on breast cancer is statistically significant.  Second, I would want to see an effect size.  The first indicates the probability that any difference among Hollywood types is due to chance, and the second indicates how big the difference is.  Both are needed.  If there is a significant difference, but the effect size shows that the risk is (as an example) 2% higher, I wouldn't say that's particularly meaningful.  Similarly, if the effect size shows that the risk is 100% higher but the p-value is .5 (ie, 50% chance that it's due to chance), I wouldn't make much of it.  On the other hand, if it was significant and the effect size was on the order of a 10% increase in risk or more, I would think of that as meaningful.  Notably, if your hypothesis is correct and there is a genuine effect, the effect size is likely to be closer to 10% higher than 100 or even 10x higher.  Cancers are mostly determined by multiple risk factors so it's unlikely that one is going to account for all that much variance.  [[User:Murray|Murray]] 12:32, 20 May 2008 (EDT)

Well, you didn't ask me the question, so I didn't realize you wanted me to answer it, but ok.  Before I do though, I will say that without a description of the method used, the reader is unable to make any informed judgment about whether there's cause for concern.  That done, there are two things that I would want to see.  First is whether the effect of Hollywood on breast cancer is statistically significant.  Second, I would want to see an effect size.  The first indicates the probability that any difference among Hollywood types is due to chance, and the second indicates how big the difference is.  Both are needed.  If there is a significant difference, but the effect size shows that the risk is (as an example) 2% higher, I wouldn't say that's particularly meaningful.  Similarly, if the effect size shows that the risk is 100% higher but the p-value is .5 (ie, 50% chance that it's due to chance), I wouldn't make much of it.  On the other hand, if it was significant and the effect size was on the order of a 10% increase in risk or more, I would think of that as meaningful.  Notably, if your hypothesis is correct and there is a genuine effect, the effect size is likely to be closer to 10% higher than 100 or even 10x higher.  Cancers are mostly determined by multiple risk factors so it's unlikely that one is going to account for all that much variance.  [[User:Murray|Murray]] 12:32, 20 May 2008 (EDT)

::Murray, don't presume ASchlafly has no evidence.  This is his site, and I'm sure he, more than anyone, is ready with verifiable evidence, just as his Commandment #1 demands.  Mr.Schlafly - I'm sure you're about to provide that evidence, and I'm sure you realize that asking rhetorical questions of your interlocutors is not the same as providing the methods behind your theory?  [[User:EngelUmpocker|EngelUmpocker]] 23:17, 19 May 2008 (EDT)

::Murray, don't presume ASchlafly has no evidence.  This is his site, and I'm sure he, more than anyone, is ready with verifiable evidence, just as his Commandment #1 demands.  Mr.Schlafly - I'm sure you're about to provide that evidence, and I'm sure you realize that asking rhetorical questions of your interlocutors is not the same as providing the methods behind your theory?  [[User:EngelUmpocker|EngelUmpocker]] 23:17, 19 May 2008 (EDT)