CODEWITHSUNDEEP
ruser-defined-functions
michael millican
michael millican

Reputation: 23

Creating a function in R to subset and run AUDPC analysis

I am trying to run an AUDPC analysis on some data. The data needs to be split into treatments and replicate blocks for the analysis. Below I have tried to write a function with the aim of spliting the data analyzing and ending up with the output of all treatment-blocks in one table/object

library("agricolae")
library("plyr")
library("dplyr")
data = read.csv("Bio2018.csv", header = TRUE, sep = ",")
data$Treatment = as.character(data$Treatment)
data$Block = as.character(data$Block)
data$Time = as.numeric(data$Time)
AUDPC.rel = function(Treatment, Block){
  data.Treatment <- subset(data, data$Treatment == Treatment & data$Block == Block)
  AUDPC = data.frame(audpc(data.Treatment$Foci, data.Treatment$Time, type = "absolute"))
}
variables = expand.grid(Treatment = c(unique(data$Treatment)), Block = c(unique(data$Block)))
Relative.AUDPC = mdply(variables,AUDPC.rel)
Relative.AUDPC

The output always give the same number for all treatment-blocks. As shown:

Treatment - Block - evaluation
Non-treated - Rep.1 - 200;
Emerald - Rep.1 - 200;
Nortica - Rep.1 - 200;
Non-treated - Rep.2 - 200;
Emerald - Rep.2 - 200;
Nortica - Rep.2 - 200;
Non-treated - Rep.3 - 200;
Emerald - Rep.3 - 200;
Nortica - Rep.3 - 200;

Upvotes: 2

Views: 228

Answers (1)

Parfait
Parfait

Reputation: 107652

Consider base R's by to subset a data frame by one or more factors such as Treatment and Block. And because by will return a list of data frames (output of function), run do.call to row bind all df elements to one final data frame.

Also, since by runs across multiple factors to split across all possible combinations, wrap a tryCatch for such special subset cases of zero rows that may raise an error of audpc call.

# BUILD FUNCTION TO RECEIVE SUBSET DF
AUDPC.rel <- function(sub) {    
  tryCatch(data.frame(Treatment = sub$Treatment[[1]],
                      Block = sub$Block[[1]],
                      Evaluation = audpc(sub$Foci, sub$Time, type = "absolute")),
           error = function(e)
                     data.frame(Treatment = sub$Treatment[[1]],
                                Block = sub$Block[[1]],
                                Evaluation = NA)
           )
}

# CALL BY(), PASSING SUBSETS TO DEFINED FUNCTION
df_list <- by(data, data[c("Treatment", "Block")], AUDPC.rel)

# BUILD SINGLE FINAL DF
final_df <- do.call(rbind, df_list)

Upvotes: 1

Related Questions