5.1 What is DiCA?


DiCA - essentially perform MCA on Sum of Observation Groups on a nominal dataset. It is working like a supervised ML algo in which it classifies observations. DiCA first creates a new space using Sum of Observation Groups, then projects the observations into that same space to see which classification would fit best.

Notes on Some Important Aids:

  1. Heat Map: This is essentially a correlation plot of summations of counts from the disjunctively coded data with rows corresponding to the group and columns corresponding to levels of variables.
  2. Fixed / Random - effect confusion matrices: This plot helps us measure accuracy of DiCA by comparing Actual VS Predicted results from the DiCA.

Author’s Notes:

Similar to BADA and PCA, DiCA is the “Supervised” Version of MCA and thus can benefits from avoiding overfitting or bias during training and improves in classification performance. The process to do DiCA is relatively easy given results from MCA. The Mathematical ideas behind DiCA is very similar to MCA and only differs in that it uses summations of counts from the disjunctively coded data for dataset.