Function uses stepwise LDA to choose best predictors for group membership, and then "regular" LDA to predict discriminant scores.
calc_ds(data, group_info)
| data | Matrix or data frame containing PC scores of specimen (rows = specimen, cols = PCs). Can also contain an additional column that defines group membership |
|---|---|
| group_info | Numeric index of column in data that indicates actual group membership, or a dataframe with two columns (col 1 = id matching the rownames in data; col 2 = group) |
Returns tibble with columns "id" and "DS". If data contained rownames, these will be saved as ids
First uses greedy.wilks to perform stepwise forward variable selection for classification, then uses selected variables in lda
data(LondonSet_scores) group_info <- LondonSet_info %>% dplyr::select(face_id, face_sex) calc_ds(data = LondonSet_scores, group_info = group_info)#> # A tibble: 102 x 2 #> id DS #> <chr> <dbl> #> 1 001 -3.28 #> 2 002 -1.85 #> 3 003 -1.92 #> 4 004 0.214 #> 5 005 0.134 #> 6 006 -1.76 #> 7 007 -2.80 #> 8 008 2.24 #> 9 009 -2.53 #> 10 010 -0.903 #> # … with 92 more rows