Title: Damned Lies for the Armchair Designer Post by: Evangolis on June 17, 2005, 03:56:49 PM Lies, Damned Lies and Statistics. I can't remember who said that, but as the son of a mathematician, I've long realized that you have to be careful with numbers. My favorite example, 100% of children born out of wedlock have mothers, therefore motherhood is clearly the root cause of illegitemacy, and thus motherhood must be stopped before it ruins our society.
Now you can do stuff like this from your own armchair. Nick Yee is working with PARC (http://www.parc.com/research/projects/PlayOn/) to put out the PlayOn Blog (http://blogs.parc.com/playon/). That's right, actual facts about server populations and statistical analysis thereof. Entirely new sets of information to misconstrue and mangle. First result: Players in guilds (http://blogs.parc.com/playon/archives/2005/06/how_does_being.html) play about the same amount as non-guilded players. Today's topic Grouping vs Soloing (http://blogs.parc.com/playon/archives/2005/06/grouping_ratio.html#comments) This Interweb stuff is so cool. Originally stole the links from a Terra Nova article (http://terranova.blogs.com/terra_nova/2005/06/longitudinal_ce.html#comments). Title: Re: Damned Lies for the Armchair Designer Post by: Shockeye on June 17, 2005, 04:16:11 PM Mark Twain popularized it, but it seems (http://su.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics) to be attributed to to Benjamin Disraeli.
Title: Re: Damned Lies for the Armchair Designer Post by: Samprimary on July 27, 2005, 02:35:33 AM Statistics are actually really cool, but many groups are negligent with the raw data, jumping to turn spurious correlations into tangible links, in cause and effect relationships that just don't exist. Like your original example.
The biggest question people should be asking with any sort of statistic is what isn't related to what, and how to root out the casual relationships. An excellent example: Imagine a California study which showed that drowning death rates were strongly correlated with the sale of ice cream: the higher the consumption of ice cream, the more deaths by drowning in that week. Of course, it doesn't take much sleuth work to find out that both variables were linked to the average overall heat - as heat rises, more people go in water, more people buy ice cream. But many people jump to links that are as absurd as assuming that the ice cream consumption was causing drowning deaths. Also famous is when cause and effect is reversed. Hypothetical example: Homicide rates go up as gun ownership goes up in a town. But either condition could be said to cause the other, as homeowners might reactively buy guns when they see news about homicides in their town rise. Title: Re: Damned Lies for the Armchair Designer Post by: HaemishM on July 27, 2005, 02:50:08 PM Too many people try to use statistics as proofs of what they want the numbers to mean, and will jiggle sample sizes and anything else to find the answers they want.
Title: Re: Damned Lies for the Armchair Designer Post by: Evangolis on July 27, 2005, 04:40:38 PM One of the lessons they pounded into me in grad school was the high likelyhood that if you go looking for a specific answer, you are likely to find it, whether it is there or not.
This thread returning to life suggests to me that you guys are feeling the heat. It broke last night here, and today was like unto heaven. Title: Re: Damned Lies for the Armchair Designer Post by: Yegolev on July 28, 2005, 01:30:37 PM I was jiggling my sample sizes last week. It certainly helped me get my point across.
Title: Re: Damned Lies for the Armchair Designer Post by: schild on July 28, 2005, 01:32:07 PM What language are you all speaking?
Title: Re: Damned Lies for the Armchair Designer Post by: Yegolev on July 28, 2005, 02:43:05 PM (http://www.intuitor.com/statistics/images/BadTes1.gif)
|