Q: what is a good option for FA with Boolean Data?

[BradKML]

Problem: Currently there is a binary data (Yes/No question dataset) that could benefit from dimensionality reduction, and be applied to feature selection and regression. The data is ves_data.csv.zip

Currently there are options for doing this:

• Boolean Matrix Factorization (to be tested https://github.com/LifanLiang/EM_BMF
• Binary Polychoric Correlation Matrix (to be tested https://github.com/inuyasha2012/pypsy
• Correlation Explanation (to be tested) https://github.com/gregversteeg/CorEx
• Positive Pointwise Mutual Information (note: only good for implicit data) https://github.com/Bollegala/svdmi
• Logistic PCA (note: native Python implementation is memory-consuming) https://github.com/brudfors/logistic-PCA-Tipping

Some information to get the data started

```python from pandas import read_csv table = read_csv('ves_data.csv') total = table[[i for i in table if ('MM01' in i and i not in [ 'MM01001','MM01BR','MM01003A','MM01003B', 'MM01003C','MM010567','MM010568','MM010569','MM010570','MM010571', 'MM010572','MM010573','MM010574','MM010575','MM010576','MM010577', 'MM010578','MM010579','MM010580','MM010581']) or i in ["PA", "GIT", "AFQT", "WAIS_BD", "WAIS_GI", 'VERAW','ARRAW','VESS','ARSS']]] total = total[[i for i in table if ('MM01' in i and i not in [ 'MM01001','MM01BR','MM01003A','MM01003B', 'MM01003C','MM010567','MM010568','MM010569','MM010570','MM010571', 'MM010572','MM010573','MM010574','MM010575','MM010576','MM010577', 'MM010578','MM010579','MM010580','MM010581']) or i in ['AFQT']]].dropna() # 'GIT' is good too from sklearn.utils import shuffle X, y = shuffle(total.drop(['AFQT'], axis=1), total['AFQT'], random_state=13) X = X - 1 # calibrating the range from 1~2 to 0~1 X = X.to_numpy() # needed for some code to function ```

[MaxHalford]

I'm closing this because I'm not sure what there is more to say. There might indeed be better methods for handling boolean data. Feel free to contribute one if you can show it's relevant.