Cute Analytics Exercise

[BeanDip4All]

meh, not even a t statistic, let alone an F test.

80 episodes is definitely enough for a T-test, t-test tables stop at 28-31 samples after that you could use the standard distribution.

I would like to know how many of 430 crew members wore red shirts, and what the percentage of the people beamed down to a planet wore red shirts.

On the whole, I am having far too much fun with this


Scotty: Cap'n, the inference engines cannae ta'e it! We need more statistical powerrr!

Kirk: Spock, analysis?

Spock: As Mr. Scott so forcefully indicates, with these small sample sizes, we are in grave danger of making a type I or type II error. Logic dictates further study.

Kirk: Mr. Sulu, plot courses for all known peaceful, female-only planets - android and alien females included!

[skiguide]

hehe.... matt is one of my good friends in the biz.. I've seen him explain this in person/in a presenation so many times, i don't need to read it.

But I can testify - it's/he's not as geeky as it sounds

I'll have to let him know you all are talking about it - he may jump in...

[Ogre]

He was wearing a red shirt.

GRRRRRRRRRRRRRRRRR!!!!!!!!!!!

[smitchell333]

That is funny:

[BeanDip4All]

hehe.... matt is one of my good friends in the biz.. I've seen him explain this in person/in a presenation so many times, i don't need to read it.

But I can testify - it's/he's not as geeky as it sounds

I'll have to let him know you all are talking about it - he may jump in...

You can tell him he could greatly improve the study by adding some p values, odds ratios, and effect sizes. Otherwise he just has post-hoc analysis and doesn't really prove any causality.

Still, it was very clever and I really enjoyed it! Also he gets major props for referring to Tufte... have you ever seen any of his books about graphs?! They're so amazing!! Any statistician would get excited by them

[BeanDip4All]

You can tell him he could greatly improve the study by adding some p values, odds ratios, and effect sizes. Otherwise he just has post-hoc analysis and doesn't really prove any causality.

and, you know what else you could tell him that we could do that would really be awesome? put together a more comprehensive set of variables and develop a predictive model!! maybe some neural networks, throw in some bayesian logic, even a markov chain could work given the right assumptions!

[13]

wow look at you go, beandip.

when you finish your current studies, serve your namesake at the party and collect samples to analyze. add beer for larger sample sizes.

ps - star wars all the way.

[woodstocksez]

you know, taking a second look at it, a cursory view of the data makes me wonder whether there weren't more red-shirt people in the first place.

in other words, 73% death-rate among Red-Shirt crewperson is telling, compared to the 8% of the Blue-Shirt crewperson only if the gap does not simply reflect a difference in the make-up of the crew.

any evidence of that?

Just what I thought when I skimmed through, but then I thought maybe it was because I just skimmed through. Not very good or interesting, really.

[cj001f]

Nobody has mentioned Google yet

[corn dog]

Just what I thought when I skimmed through, but then I thought maybe it was because I just skimmed through. Not very good or interesting, really.

Just some histograms with fancy graphics

[Tye 1on]

and, you know what else you could tell him that we could do that would really be awesome? put together a more comprehensive set of variables and develop a predictive model!! maybe some neural networks, throw in some bayesian logic, even a markov chain could work given the right assumptions!

Decafe TacoDip, decafe...

[BeanDip4All]

Just some histograms with fancy graphics

Very true. Personally, I'm way skeptical of Bailey's advice regarding chart construction: more is NOT better as a way to simplify a chart. In my experience, the big wigs prefer many, simple charts to one complex, comprehensive chart.

But then again, most big wigs I deal with probably couldn't follow that article to begin with.

[BeanDip4All]

[Tippster]

Dude. I am like seriously offended... I am *so* much way more Klingon than Vulcan!

Ewwww.....

[hutash]

... the complete range of Starships Enterprise, from NX-01 to 1701-E!

I sure hope you had to google those numbers. If you actually know them I am afraid you truly are a hopeless trecky. I would never even have been able to tell you how many there were, let alone their numbers. You scare me C.

[TeleMang]

Even if there is a correlation to red shirts getting killed... is that the causal agent responsible (as we all know; correlation is not causation)? For instance; do the shirt colors represent anything (rank for instance)? Perhaps the red shirts are the front line soldiers while the blue shirts are sitting back in the white house (i mean starship enterprise... how silly of me )fucking things up for the red shirts.