two-pot screamer dot com

From Slashdot - Firefox Continues Gains against IE

Re:And what’s the margin of error in the polling? (Score:4)
by l2718 (514756) on Saturday January 22, @12:25PM (#11441825)

Mod parent up!

\begin{rant}
Statistical figures (or any “scientific” figures, for that matter) are mostly meaningless without an error estimate (a.k.a. “confidence interval”). In fact, the lack of such estimates has been found to be a strong indication of bad research in 57.3% of all cases.

TFA claims IE market share to be “92.7%”. As parent succintly explains, that claim is clearly bogus: there are two separate percentages:

1. What they actually measured: entries to 5 specific websites over a 2-day period.
2. What they wanted to measure: IE market share.

Now there are two problems with the analysis: the first is that there is random noise in the measurement of (1). The second is that you cannot simply equate (1) with (2) without some justification. Normally you would combine the measurement errors coming from the noise and from the non-prefect correlation between (1) and (2) to give a confidence interval.

Somehow I doubt that you will find the claimed figures to even be accurate to within %1. Hence the observed rise could be entirely due to random fluctuations or other errors and is likely completely insignificant.
\end{rant}

This entry was posted on Sunday, January 23rd, 2005 at 1:12 pm and is filed under Scrapbook. You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.