BGonline.org Forums

Why to stop at multiples of 1296

Posted By: Timothy Chow
Date: Wednesday, 7 November 2012, at 12:13 a.m.

In Response To: 4-ply RO — Tips on Trial Counts (Daniel Murphy)

There is one reason to stop at multiples of 1296 that hasn't been mentioned so far in this thread.

If you stop as soon as the JSD reaches a certain value, then the statistical significance can be a little misleading. Most people will just convert the JSD to a probability using tables for the normal distribution. However, there's an important difference between

1. rolling out a predetermined number of trials, and noting that the standard deviation is x; and

2. examining the standard deviation after every single trial, and stopping as soon as the standard deviation drops below x.

The tables are based on the assumption that you're doing #1 and not #2. Intuitively, in #2, if by random chance you get a flukily low standard deviation early on, you'll halt instantly, even if continuing on might demonstrate that there's more variance left than you think.

By conventionally stopping at multiples of 1296, this side-effect of "watching your step" is minimized, and the JSDs are closer to meaning what you think they mean.

Messages In This Thread

• 61P-43U-43-d
  Robert Andersson -- Monday, 5 November 2012, at 9:19 a.m.
  • 61P-43U-43-d
    Casper van der Tak -- Monday, 5 November 2012, at 9:56 a.m.
    • 61P-43U-43-d
      konstantin -- Monday, 5 November 2012, at 2:47 p.m.
  • 4-ply RO
    Robert Andersson -- Monday, 5 November 2012, at 7:48 p.m.
    • 4-ply RO
      konstantin -- Monday, 5 November 2012, at 7:57 p.m.
      • New RO
        Robert Andersson -- Monday, 5 November 2012, at 8:19 p.m.
        • New RO
          konstantin -- Monday, 5 November 2012, at 8:30 p.m.
    • 4-ply RO — Tips on Trial Counts
      Taper_Mike -- Monday, 5 November 2012, at 9:34 p.m.
      • 4-ply RO — Tips on Trial Counts
        Stick -- Monday, 5 November 2012, at 9:43 p.m.
        • 4-ply RO — Tips on Trial Counts
          Taper_Mike -- Tuesday, 6 November 2012, at 8:51 a.m.
          • 4-ply RO — Tips on Trial Counts
            Daniel Murphy -- Tuesday, 6 November 2012, at 5:17 p.m.
            • 4-ply RO — Tips on Trial Counts
              Bob Koca -- Tuesday, 6 November 2012, at 8:47 p.m.
              • 4-ply RO — Tips on Trial Counts
                Maik Stiebler -- Wednesday, 7 November 2012, at 4:01 p.m.
                • 4-ply RO — Tips on Trial Counts
                  Bob Koca -- Wednesday, 7 November 2012, at 5:49 p.m.
                  • 4-ply RO — Tips on Trial Counts
                    Maik Stiebler -- Wednesday, 7 November 2012, at 8:09 p.m.
                    • 4-ply RO — Tips on Trial Counts
                      Daniel Murphy -- Wednesday, 7 November 2012, at 9:07 p.m.
                      • 4-ply RO — Tips on Trial Counts
                        Maik Stiebler -- Thursday, 8 November 2012, at 6:27 a.m.
            • Why to stop at multiples of 1296
              Timothy Chow -- Wednesday, 7 November 2012, at 12:13 a.m.
              • Why to stop at multiples of 1296
                Daniel Murphy -- Wednesday, 7 November 2012, at 5:44 a.m.
                • Why to stop at multiples of 1296
                  Timothy Chow -- Wednesday, 7 November 2012, at 3:30 p.m.
                  • Stepping OT now: baseball batting averages
                    Chuck Bower -- Thursday, 8 November 2012, at 2:17 p.m.
  • 61P-43U-43-d
    Caryl Stephanson -- Tuesday, 6 November 2012, at 11:31 p.m.