pca_compute() runs stats::prcomp() on a set of variables (factors and
logicals coded numerically). pca_scores_join() joins the component
scores back to the data.
Usage
pca_compute(
data,
vars,
id_col = NULL,
scale = TRUE,
impute = c("none", "median")
)
pca_scores_join(pca_obj, data, id_col = NULL, n_pc = 2)Arguments
- data
A data frame.
- vars
Variables (names or tidyselect).
- id_col
Identifier column (default: row numbers).
- scale
Scale the variables to unit variance.
- impute
"none"(complete cases only) or"median"(simple median imputation, for visualisation only, never for inference).- pca_obj
Output of
pca_compute().- n_pc
Number of components joined.
Value
pca_compute(): list with pca (prcomp object), ve (variance
explained, %), scores (tibble with .id), keep, vars, n,
impute, loadings and cor_var (correlations of the variables with the
components). pca_scores_join(): the rows of data kept in the PCA, with
the columns PC1, PC2, ...
Examples
pc <- pca_compute(cohort_sim, c(NFL_V0, IL6_V0, CRP_V0, MOCA_V0, UPDRS_V0, AGE), id_col = ID)
round(pc$ve, 1)
#> [1] 31.5 19.5 16.4 14.0 10.0 8.6
head(pca_scores_join(pc, cohort_sim, id_col = ID)[, c("ID", "PC1", "PC2")])
#> ID PC1 PC2
#> 1 P001 -2.27995419 1.0463946
#> 2 P002 0.01221179 1.4245590
#> 3 P003 1.18786860 0.9402560
#> 4 P004 1.59959226 1.2156418
#> 5 P005 0.12387455 -1.6087485
#> 6 P006 -0.57484024 0.1797266