TuningCorr_df: TuningCorr_df - Automated correction of PAM using tuned mean...
In SchoenbergA/PAMcorrection: Package for PAM correction

Description Usage Arguments Details Value Note Author(s) Examples

View source: R/TuningCorr_df.R

uses the mean based on the data clipped by 'Whisker' and 'Quantil' for specified subgroups to correct the PAM values.

1	TuningCorr_df(df, att1 = NULL, att2 = NULL, tuning = NULL, threshold = NULL)

`df`	data.frame - containing Control and PAM values with colnames "CTR" and "PAM"
`att1`	character - Column name of data frame which should be used to correct by mean. If missing, will use the global mean to correct PAM.
`att2`	character - Additional column name of data frame which should be used to correct by mean. Only necessary if the correction should use subgroups for the data in 'pos1'
`tuning`	character - method for clipping the data used to calculate the mean. Methods can be "whisker", "quantil" or "threshold". If 'NULL' will perform no tuning. Default= NULL.
`threshold`	numeric - specifies the threshold to clip values for tuning. Only needed if 'tuning' == "threshold".

tuning - available modes are "whisker" (which clips all values </> lower / upper whisker threshold in boxplot) and "quantil" (which clips all values </> 25 and 75 quantils). "threshold" will clip all values > a specified threshold.
If a specific class or combination of attributes has no values left after tuning the mean will be calculated based on the original data. This will lead to a warning. The class or combination is marked as 'noData' in the result table.
If a class or combination has a 'NaN' the class or combination is not existing in the dataframe. This is possible when using att1 and att2 an if not every att1 has all att2 classes.

returns a data.frame with an additional column "PAM_corr" with the corrected PAM values and prints the respective mean values for each class or combination of classes. Uses 'tuned' data to calculate the mean values.

If used without any attributes will use the global mean over all values to correct PAM. IF 'att1' is set will calculate the mean for each group in 'att1' to correct PAM. IF 'att2' is given will use the mean for each group with the combination for each attribute.

Andreas Schönberg

# load data
dat <- read.csv(system.file("extdata","exp_PAM.csv",package = "PAMcorrection"))
head(dat)

# how the tuning works ("whisker" and "quantil")
bp <-boxplot(dat$PAM-dat$CTR)
bp$stats[1,1] # lower whisker
bp$stats[2,1] # lower quantil
bp$stats[3,1] # median
bp$stats[4,1] # upper quantil
bp$stats[5,1] # upper whisker

# correct df by global means
corrected <-TuningCorr_df(dat,tuning = "whisker")
corrected <-TuningCorr_df(dat,tuning = "quantil")

# use one class (no tuning)
corrected <-TuningCorr_df(dat,att1 = "dilect")
# use one class with threshold tuning
corrected <-TuningCorr_df(dat,att1 = "dilect",tuning="threshold",threshold=0.8)
# use two classes (means for each generation depending on dilect) tuned by 'whisker'
corrected <-TuningCorr_df(dat,att1 = "generation",att2 = "dilect",tuning = "whisker")