Halfbakery: Statistics
Non-normal vote distribution   (+2)  [vote for, against]

From reading a limited sample of postings on the Halfbakery, there seem to be a greater proportion with votes having high overall totals (HOT) compared with those having negative overall totals (NOT). The surplus of HOT posts pulls the mean away from zero. Another feature in the data set is the existence of several, very HOT posts. These skew the distribution so that it cannot be considered normal. Statistical analysis was performed to quantify these observations. A sample size of N=100 posts were taken using the 'random' button. The data are summarised as follows:

Positive Votes
--------------------
Mean 5.80 Variance 39.3 Skewness 3.19 Kurtosis 14.3

Negative Votes
----------------------
Mean 2.73 Variance 8.74 Skewness 1.63 Kurtosis 3.31

Total Votes
-----------------
Mean 3.06 Variance 46.1 Skewness 2.24 Kurtosis 8.9

The Total Votes results show a non-zero mean with significant skewness in the positive direction, which with the high kurtosis is suggestive of a few, high valued outlier points mixed with a large number of values localised somewhat below the mean. Also of interest is the Negative Votes data, which has a comparatively small variance. Note also that the number of negative votes cast is approximately half the number of positive votes which may reflect the tendency is for large NOT posts to be excluded from the population, although may also be indicative of a general warm fuzziness that makes everyone want to vote positive.

In conclusion, there is a definite bias in the hb vote counts which is perhaps due to an inadequate 'random' function but more likely is just how it is. Therefore I propose that the croissant be replaced by a dimensionless, normalized voting parameter, lambda-pi, to give a less biased measure of an idea's quality.
-- lubbit, Nov 14 2002

But if the replacement for the croissant was dimensionless, you wouldn't be able to see it...
-- hippo, Nov 14 2002


But de lamb-da pi-a is-a beta eta, no?
-- lubbit, Nov 14 2002


Fishbones are usually displaced because:
Trolls are excommunicated, taking the fishbones with them
The idea is baked/halfbaked, then [MFD], eventually [Ð].
Bakers destroy their own *rank* ideas which seemed good at the time, or were removed to free space on the HB, regardless of bone:croissant ratio(remember that fad?)
-- thumbwax, Nov 14 2002


<mega pie> Lambda-pi: like a croissant, but smaller.<mega pie>

Also, the population you have chosen, though random, is not a representative cross section. The result will necessarily be skewed as the ideas containing most negative votes are removed from consideration. Er... in other words, what TW said.
-- egbert, Nov 14 2002


Er... Quiet day, [lubbit]?
-- whimsickle, Nov 14 2002


egbert: Also what I said, "...excluded from the population...".

However excluding the 3 highest valued points (with sums >20), which account for most of the skewness and variance, still gives a positive average. Likewise the median value, which is robust against outlying values, is 2. So the two hypotheses, that freak high values (a) skew the distribution and (b) elevate the mean above zero are, respectively, true and false. To explain the elevated mean we must hypothesise that people exclude posts which are only slightly negative, or are generally more inclined to vote positive.
-- lubbit, Nov 14 2002


We'd need a larger sample to be sure tho'.
-- lubbit, Nov 14 2002


Oh yes. <crawls back under stone>

Ooh, ooh here's another one.

<Mario>Lambda-pi, iss-a like-a da shepherd's pie, only-a younger<Mario>.
-- egbert, Nov 15 2002


Lubbit, for heaven's sake, this is just the whole neutrino-imbalance cosmic asymmetry thing manifesting itself again. Stop looking around for Dark Matter and let it be, OK ? Do you want this pseudoreality to implode ?

I put the cat in the box.... I close the box.....
-- 8th of 7, Nov 15 2002


You would.
-- madradish, Nov 16 2002



random, halfbakery