Update concering the NJTI4
See all posts in this thread below.


Christian
Published: 01112016 04:12 am
Updated: 01112016 09:38 am

I want to give an technical update regarding the progress of the NJTI4 test.
I am building it in Swedish, mainly because I have a channel where I can test about 200 new people each day, so it's easy to get statistics. Once I feel done with it it will be translated to English and published as openscience with a creativecommons license.
So the answers are on a five item Likertscale (normative).
By changing the EI factor to be about how strongly negative outcomes affect people the EI and JP factor is more orthogonal to each other.
It makes sense from a neuroscientific perspective to have the EI factor to be related to dopamine d2 receptor density in striatum and the JP factor to be about dopamine density dynamics in prefrontal cortex.
Dopamine d2 density in striatum is strongly associated with learning from negative outcomes while prefrontal dopamine density dynamics is associated with taskswithing costs, persistent working memory and interference resolution.
Also dopamine d2 density in the striatum doesn't correlate much with prefrontal dopamine dynamics so they are not physiologically dependent much.
The TF dichotomy is about individual differences in functional connectivity to the taskpositive network (TPN) vs the defaultmode network (DMN).
The SN dichotomy is about individual differences in functional connectivity to frontopolar networks.
Cronbach's alpha for each trait (I, E, N, S, T, F, J, P) is above 0.7, Cronbach's alpha for each factor (EI, SN, TF, JP) is above 0.7 persistently.
Exploratory Factor Analysis (EFA) using varimax (orthogonal) or promax (oblique) rotation both finds 4 factor loadings which overlap the designed factors almost perfectly. Pricipal Component Analysis (PCA) finds 45 components that explain up to 80% of the variance in the data and overlaps somewhat with the designed factors.
Positive eigenvalues over 1 (Kaiser's eigenvaluegreaterthanone rule (K1 or Kaiser criterion)) is distributed with a mean at 11 with standard deviation 2.




Christian
Published: 01242016 02:30 am
Updated: 01242016 02:33 am

Seems like I'm done now. Now I'm just accumulating results to form largerscale statistics.
For 500 random results:
 Cronbach's alpha for factors range 0.79  0.89 for (TF, SN, JP, EI)
 Cronbach's alpha for traits range 0.76  0.92 (T, F, S, N, J, P, E, I)
 Exploratory factor analysis (EFA) with promax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Exploratory factor analysis with varimax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Principal component analysis (PCA) finds 4 factors which somewhat (about 50%) overlap with designed factors
 7 eigenvaluesaboveone.
 pvalue for varimax EFA is 2.73E84
 To decide how many factors to extract for EFA I use a mechanistic parallell analysis




Christian
Published: 01262016 05:24 am

Now for 750 random results, statistics are roughly the same:
 Cronbach's alpha for factors range 0.79  0.89 for (TF, SN, JP, EI)
 Cronbach's alpha for traits range 0.76  0.92 (T, F, S, N, J, P, E, I)
 Exploratory factor analysis (EFA) with promax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Exploratory factor analysis with varimax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Principal component analysis (PCA) finds 4 factors which cause up to 80% of the variance and somewhat (about 50%) overlap with designed factors
 7 eigenvaluesaboveone.
 pvalue for varimax EFA is 1.2664715884659E139, ChiSquare Statistics: 1915.2, Degrees of freedom: 590
 To decide how many factors to extract for EFA I use a mechanistic parallell analysis




Christian
Published: 01272016 03:46 am

Now for 1000 random results, statistics are roughly the same, statistics seems to be stable from somewhere 150  200 results and more.
 Cronbach's alpha for factors range 0.78  0.89 for (TF, SN, JP, EI)
 Cronbach's alpha for traits range 0.75  0.92 (T, F, S, N, J, P, E, I)
 Exploratory factor analysis (EFA) with promax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Exploratory factor analysis with varimax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Principal component analysis (PCA) finds 4 factors which cause up to 80% of the variance and somewhat (about 50%) overlap with designed factors
 7 eigenvaluesaboveone.
 pvalue for varimax EFA is 2.9842396974337E201, ChiSquare Statistics: 2308.74, Degrees of freedom: 590
 To decide how many factors to extract for EFA I use a mechanistic parallell analysis




Christian
Published: 02082016 05:45 am
Updated: 02082016 05:46 am

Now for another 1000 random results, statistics are pretty much the same, statistics seems to be stable from somewhere 150  200 results and up.
 Cronbach's alpha for factors range 0.76  0.88 for (TF, SN, JP, EI)
 Cronbach's alpha for traits range 0.73  0.91 (T, F, S, N, J, P, E, I)
 Exploratory factor analysis (EFA) with promax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Exploratory factor analysis with varimax rotation finds 5 factors where 4 of them perfectly matches designed factors
 Principal component analysis (PCA) finds 4 factors which cause up to 80% of the variance and somewhat (about 50%) overlap with designed factors
 7 eigenvaluesaboveone.
 pvalue for varimax EFA is 2.8E228, ChiSquare Statistics: 2473, Degrees of freedom: 590
 To decide how many factors to extract for EFA I use a mechanistic parallell analysis




Christian
Published: 02292016 04:47 am
Updated: 02292016 04:47 am

Now the first 200 random results on the english version, statistics are pretty much the same.
 Cronbach's alpha for factors range 0.81  0.92 for (TF, SN, JP, EI)
 Cronbach's alpha for traits range 0.78  0.93 (T, F, S, N, J, P, E, I)
 Exploratory factor analysis (EFA) with promax rotation finds 4 which perfectly matches designed factors
 Exploratory factor analysis (EFA) with varimax rotation finds 4 which perfectly matches designed factors
 Principal component analysis (PCA) finds 3 factors perfectly matches 3 of the 4 designed factors
 8 eigenvaluesaboveone.
 pvalue for varimax EFA is 1.09E71, ChiSquare Statistics: 1485, Degrees of freedom: 626
 To decide how many factors to extract for EFA I use a mechanistic parallell analysis

