@reckful, you are a pleasant little contradiction, aren't you?
How about a bit of meta perspective: earlier in this thread you claimed there was a "fatal" flaw in my methodology. Now you demonstrate a willingness to move on from that point, but there is just as much confidence in your tone as before. Why, it appears as if you have unlimited confidence in your doomsday scenarios no matter how frequently they fail to turn out to be true.
Moving forward, if you don't reply directly to the content of my posts, I'm not going to bother replying to yours. That's not a conversation. It's people speaking past eachother.
User Tag List
Thread: MBTI and Big 5 correspondence

09172013, 08:43 PM #41
 Join Date
 Aug 2013
 MBTI
 ISTP
 Posts
 353

09172013, 09:15 PM #42
 Join Date
 Aug 2013
 MBTI
 ISTP
 Posts
 353
Here's a hopefully at least somewhat clarifying update on where we are. Given just the Big 5 / MBTI simple correlations, the problem of finding the regression coefficients was underdetermined (as @reckful pointed out). However, there is a simple mathematical relationship that shows us that the missing data was the Big 5 intercorrelations. So, I pulled those from another paper. Having these intercorrelations means that the system is not actually underdetermined anymore. In fact, it's fully determined. The problem is that the intercorrelations aren't from the same dataset. However, they are from the same authors, from papers that are only two years apart, and both of which have good power. This means we can trust the data.
This all also means we can in fact interpret the regression coefficients. So what are we to make of the fact that they look funny? There are three points to make. 1) We know there is error because the Big 5 does not explain all the variance in the MBTI. Far from it, in fact. 2) We introduced more error by pulling data from two different papers, even though it's technically all the data we need. 3) There could be a typo in the data entry! I always double check, but still. 4) I haven't validated the models (ie tested them on 30 mechanical turk subjects). It could be the case that the coefficients look funny, but the model still works on average. The coefficients could look funny for all sorts of reasons.

09172013, 09:21 PM #43
More smoke. Integrity much?
Your first paragraph implies that I said something earlier in the thread with "confidence" in my "tone" that I'm now acknowledging was wrong. But, on the contrary, in my secondtolast post I emphasized that my specific criticisms of your botched formulas were "assuming your overall approach here makes any theoretical sense — which, for the reasons described in my earlier posts, I don't think it does."
We're not "speaking past each other," mingularity. You're either being amazingly obtuse or annoyingly disingenuous or a charming mix of the two.
I've inarguably "replied directly to the content" of the post with your latest formula. And you have yet to explain how the 1989 and 1991 data you're citing could possibly lead to a formula in which the most important Big Five component for determining whether someone's an MBTI S or N is their Neuroticism score.

09172013, 09:26 PM #44
 Join Date
 Aug 2013
 MBTI
 ISTP
 Posts
 353
Botched formulas? Its a completely straightforward equation. Good bye.

09192013, 02:21 PM #45
 Join Date
 Aug 2013
 MBTI
 ISTP
 Posts
 353
Good news! I found a bug in my code  if you don't set byrow=TRUE in R it fills out your matrices by column first.
HTML Code:# Correlation matrix among the Big 5 predictors. From Costa, McCrae & Dye (1991) BigFive = matrix(c(1,.21,.05,.25,.49,.22,1,.43,.07,.22,.05,.43,1,.06,.04,.25,.07,.06,1,.13,.49,.22,.04,.13,1),nrow=5,ncol=5, byrow=TRUE) [N] [E] [O] [A] [C] [E] 0.22 1.00 0.43 0.07 0.22 [O] 0.05 0.43 1.00 0.06 0.04 [A] 0.25 0.07 0.06 1.00 0.13 [C] 0.49 0.22 0.04 0.13 1.00 # Mean of male / female correlations between the outcome variables (MBTI) and the predictors (Big 5). From McCrae & Costa (1989) # BigFiveMBTI = matrix(c(.17,.72,0,.06,.08,.03,.16,.71,.04,.13,.17,.15,0,.45,.19,.08,.18,.28,0,.48), nrow=4, ncol=5, byrow=TRUE) [N] [E] [O] [A] [C] [EI] 0.17 0.72 0.00 0.06 0.08 [SN] 0.03 0.16 0.71 0.04 0.13 [TF] 0.17 0.15 0.00 0.45 0.19 [JP] 0.08 0.18 0.28 0.00 0.48 # Matrix of regression coefficients for Big 5 to MBTI mapping t(solve(BigFive) %*% t(BigFiveMBTI)) [N] [E] [O] [A] [C] [EI] 0.89 1.35 0.59 0.12 0.77 [SN] 0.3 0.34 0.86 0.04 0.32 [TF] 0.38 0.32 0.13 0.56 0.01 [JP] 0.28 0.17 0.34 0.02 0.65
EI = .89N + 1.35E  .59O + .12A + .77C
SN = .3N  .34E + .86O + .04A  .32C
TF = .38N + .32E  .13O + .56A  .01C
JP = .28N  .17E + .34O + .02A  .65C
The next step is to search for a better source of Big 5 intercorrelations, or a paper that does the Big 5 / MBTI analysis *and* publishes their correlations. OR, I'm also considering collecting my own dataset on Mechanical Turk. We'll see... which random route the ISTP picks the next time he chooses this adventure ^_^
PS: There might be another bug.. in the matrix dimensions or transcription. TODO: check for this next time.

09192013, 10:35 PM #46
You cant make the distinction between T and F for men based on those correlations. Or there is some correlation, but not enough to be anything significant. S/N scale correlates with both openness and agreeableness. Agreeableness correlates with different things on men and women. All these things are enough to not be able to get full type for other than some females who score some exact numbers(and even then there is just statistical correlation, which might not fit to that person).
But what you can do to get both big 5 and MBTI scores is to use johari/nohari thing and calculate the scores for both MBTI and big 5 from those scores. It wont be as accurate as good(new official) MBTI or big 5 tests, but i read some study a while ago that its possible to get reasonably good accuracy using johari/nohari(i dont remember if the adjectives were selected for this task or if they just used some regular adjectives used in johari/nohari normally) to get scores for both tests.
Similar Threads

Big 5 is more accurate than MBTI, and is favored by researchers
By The Wailing Specter in forum Other Personality SystemsReplies: 33Last Post: 08202016, 04:05 PM 
MBTI and children
By Ivy in forum MyersBriggs and Jungian Cognitive FunctionsReplies: 30Last Post: 05102010, 01:30 PM 
MBTI and relativism
By rivercrow in forum Philosophy and SpiritualityReplies: 10Last Post: 07222007, 10:54 PM 
Web Link: MBTI and Christianity?
By Usehername in forum MyersBriggs and Jungian Cognitive FunctionsReplies: 8Last Post: 06282007, 03:25 PM 
"The Shield" MBTI and Etypes...
By The Ü™ in forum MyersBriggs and Jungian Cognitive FunctionsReplies: 2Last Post: 05272007, 06:32 AM