filter_NA: Filter features
In lucyleeow/CellProfileR: Analyse Cell Morphology Data Generated By CellProfiler

Description Usage Arguments Value Functions

View source: R/filter_features.R

Filters columns (features) of Cell Profiler data to exclude any columns with NA/NaNs, low variance or are highly correlated.

filter_NA(df, NA_cutoff, out_cols = FALSE)

filter_lowVar(df, freqCut = 95/5, uniqueCut = 10, out_cols = FALSE)

filter_cor(df, cor_cutoff, out_cols = FALSE)

`df`	Dataframe of Cell Profiler data to be filtered.
`NA_cutoff`	Optional argument. Columns with a total number of NAs or NaNs greater than the NA_cutoff number will be removed.
`out_cols`	Single logical indicating whether the filtered column names should be output as well.
`freqCut`	Passed to `caret::nearZeroVar`. The cutoff for the ratio of the most common value to the second most common value.
`uniqueCut`	Passed to `caret::nearZeroVar`. The cutoff for the percentage of distinct values out of the number of total samples.

A dataframe with the filtered columns removed. If out_cols is set to TRUE, a list of 2, where the 1st element is the filtered dataframe and the 2nd element is a vector of the filtered column names.

filter_NA: Filters columns with any NA/NaN's or a total number above a certain number if 'NA_cutoff' given.
filter_lowVar: Filters columns with low variance (e.g. if entire column consists of the same value) using the caret::nearZeroVar function. See caret documentation for details on this function.
filter_cor: Calculates the Pearson correlation matrix for the 'Median' columns then filters columns that are highly correlated to each other using caret::findCorrelation. Note that you should run filter_lowVar on your data before filtering for high correlations.