#
#
############### 'Some' useful Slicing and Dicing of Data ###############
#
#
# This script assumes you have worked through all the previous notes from 
# the web page and you have downloaded, installed, and updated all available
# R packages. 


# Slicing and dicing functions are often useful when dealing with data which is in Long 
# format and one needs the data in Wide format, or vice versus. Long format refers to 
# data in which each observation or participant has multiple rows. Wide format refers to 
# data in which each observation or participant has only one row. 

# Creating a data file for an example of extracting specific elements from a data file. 

x1.a <- rep(1, 10)
x1.b <- rep(2, 10)
x1.c <- rep(3, 10)
x1.d <- rep(4, 10)
x1.e <- rep(5, 10)
x1.f <- rep(6, 10)
x1.g <- rep(7, 10)
x1.h <- rep(8, 10)
x1.i <- rep(9, 10)
x1.j <- rep(10,10)
participant.id <- c(x1.a, x1.b, x1.c, x1.d, x1.e, x1.f, x1.g, x1.h, x1.i, x1.j)

x1.numbers <- rep(seq(1:10), 10)
x1.let <- factor(LETTERS[1:10]); names(x1.let) <- letters[1:10]
x1.letters <- rep(x1.let, 10)

x2.numbers <- rep(seq(1:5), 20)
x2.let <- factor(LETTERS[1:5]); names(x2.let) <- letters[1:5]
x2.letters <- rep(x2.let, 20)

x3.numbers <- rep(seq(1:2), 50)
x3.let <- factor(LETTERS[1:2]); names(x3.let) <- letters[1:2]
x3.letters <- rep(x3.let, 50)

x4 <- rnorm(100)
x5 <- rnorm(100, 100, 15)
x6 <- rnorm(100, 5, 1)

WL.data <- data.frame(participant.id, x1.numbers, x1.letters, x2.numbers, x2.letters, x3.numbers, x3.letters, 
                      x4, x5, x6)

# Cleaning up the workspace (house cleaning) by removing all the things we created to 
# make the data.frame.

ls()
rm(participant.id,x1.a,x1.b,x1.c,x1.d,x1.e,x1.f,x1.g,x1.h,x1.i,x1.j,x1.let,x1.letters,x1.numbers,
   x2.let,x2.letters,x2.numbers,x3.let,x3.letters,x3.numbers,x4,x5,x6)
ls()

#####################################################################################

# Take a look at the created data

WL.data

head(WL.data, n = 10)


# The following can identify individuals with value "B" on variable "x2.letter" of data frame "WL.data".

subset.1 <- which(WL.data$x2.letter == "B")
subset.1

# The following gives a summary of all columns (variables) by "x2.letters".

subset.2 <- by(WL.data[ , 1:10], WL.data$x2.letters, summary)
subset.2

names(subset.2)

subset.2$B

# Creates a new data frame with just the participant's row data for the rows 
# having a "C" on "x2.letters" in "WL.data".

subset.3 <- data.frame(WL.data[which(WL.data$x2.letter == "C"), 1:10])
subset.3

# Creates a new data frame with just the participant's row data for the rows 
# having a "B" on "x3.letters" in "WL.data".

subset.4 <- data.frame(WL.data[which(WL.data$x3.letter == "B"), 1:10])
subset.4

# Create a new data frame with just the participant's row data for the rows 
# having a "C" on "x2.letters" in "WL.data" and the columns of the new data frame 
# include x4, x5, x6 (which are columns 8, 9, 10 in WL.data).

subset.5 <- data.frame(WL.data[which(WL.data$x2.letter == "C"), 8:10])
subset.5

# Same as above (subset.5), accept the new data frame contains participant.id (column 1) as well.

subset.5a <- data.frame(WL.data[which(WL.data$x2.letter == "C"), c(1,8:10)])
subset.5a

# Create a new data frame with just the participant's row data for the rows 
# having a "B" on "x3.letters" in "WL.data" and the columns of the new data frame 
# include pariticipant.id, x4, and x6 (which are columns 1, 8, and 10 in WL.data).

subset.6 <- data.frame(WL.data[which(WL.data$x3.letter == "B"), c(1,8,10)])
subset.6

###########################################################################

### It can often take three or more steps to get a total score for a participant.

## Extract all data for just participant 1. 

p1 <- data.frame(WL.data[which(WL.data$participant.id == "1"), 1:10])
p1

## Extract just the rows which have values of B on x3.letters for participant 1.

p1.x3B <- data.frame(p1[which(p1$x3.letters == "B"), 1:10])
p1.x3B

## Create a total score for variable x4 for just participant 1 for just the 
## rows with values of B on x3.letters.

p1.x3B.x4total <- sum(p1.x3B$x4)
p1.x3B.x4total

## Or, create a total score for the sum of x4, x5, x6 at just the row for the value 
## of "H" on x1.letters. 

p1.x3B.x1H.456total <- sum(p1.x3B[1,8], p1.x3B[1,9], p1.x3B[1,10])
p1.x3B.x1H.456total

# Same as above.

p1.x3B.x1H.456total2 <- p1.x3B[1,8] + p1.x3B[1,9] + p1.x3B[1,10]
p1.x3B.x1H.456total2


#############################################################################

# Creates a data frame with the sum of "x4" for each value of "x3.letters".

subset.7 <- aggregate(WL.data[,c("x4"), drop=FALSE], 
  by=list(x3.letters=WL.data$x3.letters), FUN=sum)
subset.7

# Creates a data frame with the sum of "x4" for each value of "participant.id".

subset.8 <- aggregate(WL.data[,c("x4"), drop=FALSE], 
  by=list(participant.id=WL.data$participant.id), FUN=sum)
subset.8

# Creates a data frame with the sum of "x4" for each value of "participant.id" by each 
# value of "x3.letters". 

subset.9 <- aggregate(WL.data[,c("x4"), drop=FALSE], 
  by=list(participant.id=WL.data$participant.id, x3.letters=WL.data$x3.letters), 
  FUN=sum)
subset.9

# Recall, each participant has 5 "A" and 5 "B" (on x3.letters).

head(WL.data, n = 22)

#############################################################################

# Creates a Long data frame with the sum of "x6" for each "participant.id" by each 
# value of "x3.letters" (A & B). 

subset.10L <- aggregate(WL.data[,c("x6"), drop=FALSE], 
  by=list(participant.id=WL.data$participant.id, x3.letters=WL.data$x3.letters), 
  FUN=sum)
subset.10L

# Re-format the data set "subset.10" changing the format from Long to Wide. 

subset.10W <- data.frame(subset.10L[1:10, ], subset.10L[11:20, ])
subset.10W

mean(subset.10W$x6)
mean(subset.10W$x6.1)

#########################################################################################
#########################################################################################

############### Long to Wide and Wide to Long ###############

# The WL.data file is originally in long format. 

head(WL.data)

# Creating a new data file which removes the unnecessary columns between x1.letters and x3.letters. 

WL.data.1 <- WL.data[ , -c(3:7)]
head(WL.data.1)

###### Going from Long format to Wide format. 

# The following creates a data.frame in which each row is one participant with each x4, x5, x6 score 
# for that participant at each x1.numbers. Recall, each participant has 10 values of x1.numbers (and 
# therefore, 10 x1.letters); so, each participant in the newly created data frame has 10 columns for x4 
# scores, 10 columns for x5 scores, and 10 columns for x6 scores. This preserves all the data from the 
# original "WL.data.1" file.

wide.x1 <- reshape(WL.data.1, v.names = c("x4","x5","x6"), idvar = "participant.id",
                     timevar = "x1.numbers", direction = "wide")
head(wide.x1)

# The 'wide.x1' data file now allows us to deal with the x4, x5, or x6 scores for each participant, by 
# each value of x1.numbers we want. For instance, if we wanted the difference in x5 scores 
# between the first and last values of x2 for each participant; then we would use the following; which 
# returns a vector of 10 values, one value for each participant which represents the difference left from 
# subtracting each participant's x5 score (at time 1) from each participant's x5 score (at time 10).

diff <- wide.x1$x5.1 - wide.x1$x5.10
diff

##### Going from Wide format to Long format.

# Going from wide to long is easy because all the attributes of the data are still available 
# from the first reshape.

long.x1 <- reshape(wide.x1)
head(long.x1)

# Original 'WL.data' example.

wide2.x1 <- reshape(WL.data, v.names = c("x4","x5","x6"), drop = c("x1.letters","x2.numbers","x2.letters",
                    "x3.numbers","x3.letters"),idvar = "participant.id", timevar = "x1.numbers", direction = "wide")
head(wide2.x1)

long2.x1 <- reshape(wide2.x1)
head(long2.x1)

#
# For a better understanding of identifying or extracting elements of a vector, matrix, or list; 
# see the section "Slicing and extracting data" in the R Reference Card (Short, 2004) posted on the web site. 
# For a better understanding of the 'reshape' function and 
# all of its arguments:

help(reshape)

# It may also be helpful to consult the doBy package. 
library(doBy)

help(doBy)