R package to generate summary statistics for PRSmix elastic net
This package defines a sumstats R object. These objects are created by the make_sumstats function, which
takes two inputs: a design matrix X and an outcome vector Y. make_sumstats returns a sumstats object,
which is a list with two elements; X'X and X'Y. Attributes provide additional information: nsubj
(number of rows of X), nobs (observations with no missing values), nmiss (observations with missing values,
which are dropped before computation of summary stats), colsum (sum over columns of X), ysum (sum over Y,
this is the number of cases for binary outcomes), yssq (sum of squares of Y), centered (boolean for
whether the summary stats were centered on the mean).
make_sumstats_clusters is a convenience function that performs matching on its inputs by sample_id
(assumed to be the first column of all input data frames), removes missing values, and calls
make_sumstats on all samples together and by cluster (defined in the clusters input).
combine_sumstats combines multiple sumstats objects. The resulting object is centered and scaled.
glmnet_sumstats runs an elastic net analysis on a sumstats object.
sim_sumstats and eval_sim simulate replicates of the training data and use the output of multiple calls to
glmnet_sumstats to find the model that minimizes the loss.
More details on the methods implemented in this package can be found here.