IE value distribution
Time Elapse and Outliers
Boxplot 1
Boxplot 2
Boxplot 3
Boxplot 4
Boxplot 5
Correlation Matrix
Volcano Plot for Questions for extrovert determination
Volcano Plot for Questions for introvert determination
Conclusion

Inspiration and What it does

Our basic step is to firstly determine the most significant questions to discriminate introvert/extrovert, so we thought of conducting multiple logistic regressions, which is exactly the idea undergone in GWAS (Genome-wide association study) study: initiated a large amount of simple regressions to find the significant SNPs(Single Nucleotide Polymorphism) in determination of diseases. So we generate our idea from GWAS study to identify some key questions. And then we intend to conduct logistic regression to generate a model predicting introvert/extrovert with those key questions.

How we built it

Data processing

filter out IE = 0 (missing value)
filter out testelapse value in the top or bottom 1% (remove outliers in people spend too little or too much time on the test)

Specify variables of interest

independent variables: answers for 91 questions (categorical,1,2,3,4,5)
dependent variables: introvert/extrovert (categorical, 1= yes, 0 = no/not sure)

Data visualisation

Boxplot for values of each question
Heatmap for correlation matrix

Analysis plan

Split data to testing data and training data (30%:70%)
Select exposures using idea of GWAS(multiple logistic regression) for extrovert and introvert with threshold p=0.05/91 for multi comparison
Volcano plot with x axis=coef and y axis = -log10(pval)
Choose significant questions (ideally about 5 questions)
Logistic regression: extrovert/introvert ~ Q1+Q2+Q3+age+sex
Cross validation: Original Models , Parallel Models, and Series Models.

Challenges we ran into

Multiple testing coding and plot several variables in one plot are new to us. We searched online and figure out the appropriate and effective way to do it.

Accomplishments that we're proud of

We used the idea of GWAS which is pretty innovative.

What we learned

We leaned how to do good plots and how to do cross validation.

What's next for Introvert or Extrovert ? That is not a question!

We still touched little upon other prediction models except for logistic regression model, and we may need some time to figure them out in the future.