@dreev @thatgirl Okay, I’m now convinced that @GetReporter needs an upgrade on its random algorithm. Thanks for making me suspicious.
1
@anomalily @dreev @thatgirl what was the issue you discovered?
1
@GetReporter @anomalily From prev discussion sounded like you decided not to classify it as a bug:
Beeminder @bmndr
9 Feb 2014
Replying to @bmndr
@GetReporter @aaronpk Why uniform randomness breaks unpredictability: If no prompt in last hour (or whatever freq) then you know one's due.
1
@bmndr @anomalily That’s true. It’s a trade-off between true randomness and randomness that seems broken.
2
@GetReporter @bmndr @anomalily Actually the current randomness does seem quite broken. I'm often pinged literally... aaron.pk/r4Xq1
2
@aaronpk @GetReporter @bmndr My data reflects “breakfast” with a higher frequency than other daily activities because of report timing.
1
@anomalily Do you not make manual reports ever? I usually make more manual reports than prompted ones. @GetReporter @bmndr
1
@aaronpk @GetReporter @bmndr Never! The Point for me is random samples - best statistical data practice.
1
1
Replying to @anomalily
@anomalily @GetReporter @bmndr I lean heavily more towards the "complete dataset" side of things, which has its own set of problems.

Aug 28, 2014 · 6:10 PM UTC

1
Replying to @aaronpk
@aaronpk @GetReporter @bmndr Oh I know you do! I view reporter as a compliment using random sampling. I have many other complete datasets.
1
@aaronpk @GetReporter @bmndr It’s the closest thing to being able to do double-blind random trials on your self-tracking.