Milouse seems to know what he's doing with stats - I know he doesn't publish any he thinks are inaccurate. I don't know what method he uses for sampling, but I guess that the 5% he polls is a different 5% each time - and it keeps coming up with about the same figure
But we have to be careful not to read more into them than we should! While I believe them to be accurate statistics, I don't think they always give us the information we need to be able to judge certain things accurately. Take DR for example. The stats shows masses of players at DR D, but they aren't really the ones we're interested in because that includes many that have given up (after maybe not many more the 5 races cut-off for that stat). The information we'd like is a stat based on currently active players in some way, but 'active' is hard to define and detect. I might technically be in the top 1%, but it doesn't mean much when a high % have stopped playing!
That said, the sport mode participation stat is more straightforward. Sure, there's a bunch of beta-only players who drag it down, but then more recent buyers might push it up (by about the same?). So I think it's probably not even across all players - the conjecture is that the more recently someone acquired GTS, the more likely they are to participate in online races - but the statistics we have don't tell us whether or how those things correlate. To be clear I expect it's only a minor effect that makes little difference to the stat, but it would be interesting to know all the same.