chess.com vs FIDE(USCF) rating

There must of course be a statistically significant positive correlation between chess.com ratings and USCF/FIDE ratings though I assume the correlation is fairly weak, especially for online chess due to reasons already mentioned.

It would be interesting if members could input their FIDE/USCF ratings here, and a regression analysis could be used to calculate estimated ratings and also provide a 95% confidence interval, similar to what chesstempo.com does with their tactics training. Shouldn't be too hard to implement. I would guess that standard live chess ratings would have the smallest confidence interval and online chess the largest (meaning that standard live chess ratings would be the most accurate in predicting "real" rating).