Personality and Popularity in the Speakeasy (A Study!)
- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
Try doing polynomial regression on a few indicators maybe.In post 131, Psyche wrote:multiple regression is extremely easu using r i really should have done it near the very beginning of my data analysis
i think in general out of laziness of a lack of imagination i've stuck too closely to basic linear analyses when there's a lot more really cool statistics i can try out
All you have to do is center the variable, then add a regressor that's the square (or cube, etc., but it gets hard to interpret after cubes) of the centered indicator. The polynomial is important if the squared regressor is significant.
With personality indicators for some variables there's a "sweet spot" that's ideal whereas the extremes are pathologic. If that's true about popularity, then at least testing a polynomial regression model is probably really important for model specification.- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
I'd really like the SE 6000 word thing you wrote (or just the words that it grabs for me). I wanted to do the IBM thing before, but I don't really use Twitter or anything, so it's hard to get 6000 words that aren't theses.- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
Speakeasy has a thread whose sole purpose is as a dumping ground for angst, so it makes sense that depression would be overrepresented.- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
All the more reason.In post 149, PJ. wrote:
I'd argue that's the sole purpose of the entire speakeasy.In post 148, Iecerint wrote:Speakeasy has a thread whose sole purpose is as a dumping ground for angst, so it makes sense that depression would be overrepresented.- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
Big 5 maps pretty OK onto MB tbh. Depends on what you consider "good enough"In post 168, Elbirn wrote:*looks up*
Seems legit
Idk I think it'd be interesting to see who gets what results on Meyers Briggs, and then apply that to mafia and each persons playstyle. So like... I'm INFP (T), whatever that added T is supposed to mean. And that certainly seems to corrolate with me being an oddball and being completely illogical and evaluating based on emotion and reading tea leaves?
Idk
Psyche I'm sorry for offtopic-ing your thread, your personality test is better pls continue
Extraversion is E>I obviously.
Agreeableness is T<F.
Conscientiousness is J>P (with a loading on T>F too).
Openness is S<N (but also with small loadings on everything).
Neuroticism doesn't map to anything.- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
Yeah, factor analysis would be even better.In post 186, Psyche wrote:so i actually told my professor about this and he suggested factor analysis
he also said this stuff was publishable (though not in a journal of psychology) but maybe he wouldn't have if he knew all the ethical issues surrounding how i got my data hahahahahaha
Basically if regression is y = mx + b in algebra, structural modeling is when you did systems of equations in algebra.- Iecerint
-
Iecerint Survivor
- Iecerint
- Survivor
- Survivor
- Posts: 15766
- Joined: May 13, 2009
- Location: San Francisco
No, multiple regression is when you have likeIn post 209, Psyche wrote:wouldn't that be multiple regression
y = mx + nz + b
But they start using beta for all the coefficients to make it less confusing because they're not ordered.
But it's all one equation. Structural modeling is like:
Y = mx + b
x = betaD + betaE + betacoeff
You can make equations that predict the predictors.
Factor analysis can either be totally data driven in which case it's just principal component analysis plus a rotation (which is a silly math toy) or theoretically specified and hypothesis-tested, in which case you do something conceptually like the above. - Iecerint
- Iecerint
- Iecerint
- Iecerint
- Iecerint
- Iecerint
- Iecerint
- Iecerint