Dynamic model updating (DMU) approach for statistical learning model building with missing data

. 2021 Apr 29;22:221. doi: 10.1186/s12859-021-04138-z

Algorithm: Dynamic Model Updating

Slice the original dataset D with p features to create set d of k datasets, such that:

$d_{k} h a s n o m i s s i n g v a l u e s,$

$d_{k} \subset D,$

$⋂ (d_{l}, d_{m}) = \emptyset | d_{l}, d_{m} \in d a n d l \neq m$

Sequentially perform Bayesian Regression on each dataset k to get posterior estimate, {

{\hat{β}}_{j} | j = \{1, \dots, p\}

. The posterior estimate of

{\hat{β}}_{j}

after k-1th dataset is used as the prior estimate for kth dataset