CODEWITHSUNDEEP
rcountaggregaterows
user2983485
user2983485

Reputation: 55

Frequency of categorical values for different variables/cols in R?

I try to aggregate data from an signal detection experiment to compute the hit rate, false alarm rate etc.

   Code Cond bf1 bf2 bf3 bf4 bm1 bm2 bm3 bm4
BAX-011    3  CR  FA HIT  FA  FR  CR  FA  FA

My variables bf1 to bm3 are factors with the levels (hit,fa,cr,fr).

I want to compute the amount of hits, fa's ... for each participant (row) but with subsets of variables (bf-items and bm-items). What's the easiest way to perform that?

It should look like that in the end:

   Code Cond bf1 bf2 bf3 bf4 bm1 bm2 bm3 bm4 bf_hits bm_hits bf_fa ...
BAX-011    3  CR  FA HIT  FA  FR  CR  FA  FA       1       0     2 ...

Upvotes: 0

Views: 433

Answers (2)

A5C1D2H2I1M1N2O1R2T1
A5C1D2H2I1M1N2O1R2T1

Reputation: 193517

If I understand your question correctly, you probably just need to explore melt and dcast from the "reshape2" package. Using @zx8754's sample data, try the following:

library(reshape2)

### Make the data into a "long" format
dfL <- melt(df, id.vars=c("Code", "Cond"))

### Split the existing "variable" column. 
### Here's one way to do that.
dfL <- cbind(dfL, setNames(
  do.call(rbind.data.frame, strsplit(
    as.character(dfL$variable), "(?=\\d)", perl=TRUE)), 
  c("var", "time")))

### This is what the data now look like.
head(dfL)
#      Code Cond variable value var time
# 1 BAX-011    3      bf1    CR  bf    1
# 2 BAX-012    3      bf1    CR  bf    1
# 3 BAX-013    3      bf1    CR  bf    1
# 4 BAX-011    3      bf2    FA  bf    2
# 5 BAX-012    3      bf2    FA  bf    2
# 6 BAX-013    3      bf2   HIT  bf    2

### Use `dcast` to aggregate the data. 
### The default function is "length" which is what you're looking for.
dcast(dfL, Code + Cond ~ var + value, value.var="value")
# Aggregation function missing: defaulting to length
#      Code Cond bf_CR bf_FA bf_HIT bm_CR bm_FA bm_FR bm_HIT
# 1 BAX-011    3     1     2      1     1     2     1      0
# 2 BAX-012    3     1     2      1     0     2     1      1
# 3 BAX-013    3     1     1      2     0     2     1      1

From there, you can always merge or cbind the relevant columns together to get the full data.frame.

Update

To avoid being seen as a "reshape2" fanboy, here's a base R approach. I hope it also shows why I went the "reshape2" route in this case:

X <- grep("^bf|^bm", names(df))
df[X] <- lapply(df[X], as.character)
dfL <- cbind(dfL, setNames(
  do.call(rbind.data.frame, strsplit(
    as.character(dfL$ind), "(?=\\d)", perl=TRUE)),
  c("var", "time")))
dfL$X <- paste(dfL$var, dfL$values, sep ="_")
dfA <- aggregate(values ~ Code + Cond + X, dfL, length)
reshape(dfA, direction = "wide", idvar=c("Code", "Cond"), timevar="X")

Upvotes: 1

zx8754
zx8754

Reputation: 56004

Try this:

#dummy data
df <- read.table(text="
Code Cond bf1 bf2 bf3 bf4 bm1 bm2 bm3 bm4
BAX-011    3  CR  FA HIT  FA  FR  CR  FA  FA
BAX-012    3  CR  FA HIT  FA  FR  HIT  FA  FA
BAX-013    3  CR  HIT HIT  FA  FR  HIT  FA  FA
", header=TRUE)

#count HITs per bf bm
df$bf_hit <- rowSums(df[,colnames(df)[grepl("bf",colnames(df))]]=="HIT")
df$bm_hit <- rowSums(df[,colnames(df)[grepl("bm",colnames(df))]]=="HIT")

#output
df
#Code Cond bf1 bf2 bf3 bf4 bm1 bm2 bm3 bm4 bf_hit bm_hit
#1 BAX-011    3  CR  FA HIT  FA  FR  CR  FA  FA      1      0
#2 BAX-012    3  CR  FA HIT  FA  FR HIT  FA  FA      1      1
#3 BAX-013    3  CR HIT HIT  FA  FR HIT  FA  FA      2      1

Upvotes: 0

Related Questions