从宽格式到长格式重塑时间序列数据

Question

Given time series data of four clients, Client 1,2,3,4.

names(data)
"Date" "X1.CLIENT"   "X1seasonal"  "X1trend"  "X1remainder" "X2.CLIENT" 
  "X2seasonal"  "X2trend"     "X2remainder" "X3.CLIENT"   "X3seasonal"  
"X3trend"     "X3remainder" "X4.CLIENT"   "X4seasonal"  "X4trend"     
"X4remainder"

Please note that Every Client data is followed by the seasonal, trend and remainder components for the same period of time.

I would like to reshape into long format in a way that looks like

"Date" "CLIENT_Number" "Type"  "Value"

[Client_Number : 1,2,3,4 ; Type: Client, seasonal, trend, remainder ]

(Additional help to understand : If data(wide format) has 30 rows/instances then we have to convert into long format(30*4*4= 480 rows/instances)

Answer 1

Looks like your data is an awkward repeated wide format. As ssdecontrol says, use melt from reshape2, I would combine with something like lapply to split into sections. eg:

# Set up
install.packages("reshape2")
library(reshape2)

# Dummy data
d = data.frame(date=seq.Date(as.Date("2015-01-01"), length.out=100, by=1),
Client.1="A", trend.1=rnorm(100), remainder.1=rpois(10, 3), Client.2="B",
trend.2=rnorm(100), remainder.2=rpois(10, 3), Client.3="C", trend.3=rnorm(100),
remainder.3=rpois(10, 3))

# Split data by columns with client numbers
d = lapply(c(2,5,8), function(i){
    # Take columns relating to client i
    x = d[,c(1, seq(i, length.out=3, by=1))]
    # Rename so you have consistent factor labels
    names(x)[2:4] = c("client", "trend", "remainder")
    # Conververt from wide to long
    x = melt(x, id.vars=c("date", "client"))
})

# Make list into a data frame
d = do.call("rbind", d)

I find this post really clear about wide and long formats: http://seananderson.ca/2013/10/19/reshape.html