Complex data transformation - database

Question

I need to transform following (simplified) dataset, created by following code:
structure(list(W1.1 = structure(c(1L, NA, NA), .Names = c("case1",
"case2", "case3"), .Label = "1", class = "factor"), R1.1 = structure(c(1L,
NA, NA), .Names = c("case1", "case2", "case3"), .Label = "2", class = "factor"),
W1.2 = structure(c(NA, 1L, NA), .Names = c("case1", "case2",
"case3"), .Label = "1", class = "factor"), R1.2 = structure(c(NA,
1L, NA), .Names = c("case1", "case2", "case3"), .Label = "1", class = "factor"),
W2.1 = structure(c(NA, 1L, NA), .Names = c("case1", "case2",
"case3"), .Label = "1", class = "factor"), R2.1 = structure(c(NA,
1L, NA), .Names = c("case1", "case2", "case3"), .Label = "1", class = "factor"),
W2.2 = structure(c(1L, NA, NA), .Names = c("case1", "case2",
"case3"), .Label = "2", class = "factor"), R2.2 = structure(c(1L,
NA, NA), .Names = c("case1", "case2", "case3"), .Label = "1", class = "factor"),
W3.1 = structure(c(1L, NA, NA), .Names = c("case1", "case2",
"case3"), .Label = "1", class = "factor"), R3.1 = structure(c(1L,
NA, NA), .Names = c("case1", "case2", "case3"), .Label = "1", class = "factor"),
W3.2 = structure(c(1L, 1L, NA), .Names = c("case1", "case2",
"case3"), .Label = "1", class = "factor"), R3.2 = structure(c(1L,
1L, NA), .Names = c("case1", "case2", "case3"), .Label = "1", class = "factor"),
age = structure(c(3L, 1L, 2L), .Names = c("case1", "case2",
"case3"), .Label = c("20", "48", "56"), class = "factor"),
gender = structure(c(2L, 1L, 2L), .Names = c("case1", "case2",
"case3"), .Label = c("female", "male"), class = "factor")), .Names = c("W1.1",
"R1.1", "W1.2", "R1.2", "W2.1", "R2.1", "W2.2", "R2.2", "W3.1",
"R3.1", "W3.2", "R3.2", "age", "gender"), row.names = c(NA, 3L
), class = "data.frame")
For the new data I want:
- a row dedicated to every x.x, with info on the Rx.x value, age and gender.
- only have a row returned when Wx.x was 1. When 2 or NA, I don't need it.
For my example this dataset should look something like this:
incident type Where Reported age gender
1 1 1.1 1 2 56 male
2 2 3.1 1 1 56 male
3 3 3.2 1 1 56 male
4 4 1.2 1 1 20 female
5 5 2.1 1 1 20 female
6 6 3.2 1 1 20 female
Note: the "Where" column can even be omitted since it should be a constant vector of 1, and I don't need it for the analysis.

Answer 1

This is (mostly) a problem to be tackled by reshape(). Assuming your original dataset is called "temp":
First, reshape it from a wide format to a long format.
temp.long <- reshape(temp, direction = "long",
idvar=c("age", "gender"),
varying = which(!names(temp) %in% c("age", "gender")),
sep = "")
temp.long
# age gender time W R
# 56.male.1.1 56 male 1.1 1 2
# 20.female.1.1 20 female 1.1 <NA> <NA>
# 48.male.1.1 48 male 1.1 <NA> <NA>
# 56.male.1.2 56 male 1.2 <NA> <NA>
# 20.female.1.2 20 female 1.2 1 1
# 48.male.1.2 48 male 1.2 <NA> <NA>
# 56.male.2.1 56 male 2.1 <NA> <NA>
# 20.female.2.1 20 female 2.1 1 1
# 48.male.2.1 48 male 2.1 <NA> <NA>
# 56.male.2.2 56 male 2.2 2 1
# 20.female.2.2 20 female 2.2 <NA> <NA>
# 48.male.2.2 48 male 2.2 <NA> <NA>
# 56.male.3.1 56 male 3.1 1 1
# 20.female.3.1 20 female 3.1 <NA> <NA>
# 48.male.3.1 48 male 3.1 <NA> <NA>
# 56.male.3.2 56 male 3.2 1 1
# 20.female.3.2 20 female 3.2 1 1
# 48.male.3.2 48 male 3.2 <NA> <NA>
Second, do some cleanup.
temp.long <- na.omit(temp.long)
temp.long <- temp.long[-which(temp.long$W == 2), ]
temp.long <- temp.long[order(rev(temp.long$gender), temp.long$time), ]
rownames(temp.long) <- NULL
temp.long$incident <- seq(nrow(temp.long))
temp.long
# age gender time W R incident
# 1 56 male 1.1 1 2 1
# 2 56 male 3.1 1 1 2
# 3 56 male 3.2 1 1 3
# 4 20 female 1.2 1 1 4
# 5 20 female 2.1 1 1 5
# 6 20 female 3.2 1 1 6
You can do further cleanup to change your column names and column order if it's important.