TheRiov wrote:
I'm also making no assumptions that 100% of anyone's posts are political. Certainly 100% of my posts in hellfire are not.
Your weighted average includes 100% of the respondents' Hellfire posts. Mathematically speaking, this is exactly the assumption you are making when you assert that your results have validity.
TheRiov wrote:
But short of a more detailed analysis which no one will perform, this is as good an approximation as we're going to get.
It would be equally true to say that it's the
worst approximation we're going to get (supposing that you're right that no one will ever perform another analysis). Ordinality is meaningless in a set of 1.
TheRiov wrote:
So unless you can provide better numbers, using the methods you describe, we'll go with my analysis. You can whine about why my numbers are not 100% accurate (and again, something I freely admit) but they're a reasonable approximation given the data we have, and our analysis capabilities.
I don't see any convincing arguments that your analysis is a reasonable approximation. Several major issues have been raised about your methodology and about the methodology of the test upon which your data relies in the first place. You haven't answered any of those issues, that I can tell. The only justification you've given is that "these are the only numbers we have". That really isn't relevant. There is no particular reason, let alone obligation, to "go with" any analysis at all. If the only analysis we have is too flawed to draw meaningful conclusions, then we should still reject that analysis.
But setting aside issues with methodology and data, I'm not convinced that your numbers are even the best approximation possible using said methods and data. You've ignored an entire dimension of the data for no discernible reason.