CODEWITHSUNDEEP
rxtszoo
gabriel
gabriel

Reputation: 399

How to specify data frame as time series?

I am loading in time-series data into R for analysis. I am trying to lag one of the variables in order to difference the series. Unfortunately, the values of the differences variables all equal 0, because R wasn't successful at lagging the weight variable. I know I am supposed to use the as.ts(data$date) to specify that that "date" variable is a time series but every time I do so it changes the "date" variable into numeric numbers. Not to mention I thought I specified that the "date" column in the dataset was a time/date variable when I initially loaded it. How can I specify the data.frame as a time series?

data=read.csv("filelocation",header=T,colClasses=c("Date","numeric")

  date   weight
2010-10-04 52495    
2010-10-01 53000    
2010-09-30 52916    
2010-09-29 52785    
2010-09-28 53348    
2010-09-27 52885    
2010-09-24 52174    
2010-09-23 51461    
2010-09-22 51286    
2010-09-21 50968    
2010-09-20 49250

data=data[order(data$date),]
diffweight1=weight-lag(weight,1)

Upvotes: 3

Views: 1677

Answers (4)

G. Grothendieck
G. Grothendieck

Reputation: 269371

Try this:

library(zoo)

z <- read.zoo("filelocation", header = TRUE, sep = ",")
diff(z)

Upvotes: 4

agstudy
agstudy

Reputation: 121568

when you manipulate times series it is better to use (zoo or xts) packages. Many time series operations as lags, diff become very simple.

here an example using xts package ( I prefer this one)

# I read your data
dat <- read.table (text = 'date   weight
2010-10-04 52495    
2010-10-01 53000    
2010-09-30 52916    
2010-09-29 52785    
2010-09-28 53348    
2010-09-27 52885    
2010-09-24 52174    
2010-09-23 51461    
2010-09-22 51286    
2010-09-21 50968    
2010-09-20 49250',header=TRUE)
# I construct my xts object
dat.xts <- xts(dat$weight,order.by=as.POSIXct(dat$date))
# new 2 columns withs lags(1) and diff

merge(dat.xts, ll = lag(dat.xts),dd =diff(dat.xts))
           dat.xts    ll   dd
2010-09-20   49250    NA   NA
2010-09-21   50968 49250 1718
2010-09-22   51286 50968  318
2010-09-23   51461 51286  175
2010-09-24   52174 51461  713
2010-09-27   52885 52174  711
2010-09-28   53348 52885  463
2010-09-29   52785 53348 -563
2010-09-30   52916 52785  131
2010-10-01   53000 52916   84
2010-10-04   52495 53000 -505

Upvotes: 3

Blue Magister
Blue Magister

Reputation: 13363

Time-series objects are designed to track data sampled at equally spaced points in time. You have an uneven sampling interval, but ts(data) seems to do what you're looking for.

Upvotes: 1

Utkarsh Naiknaware
Utkarsh Naiknaware

Reputation: 126

What I feel you need is difference between adjacent rows for weight col You can try :

weight <- c(20,40,70,110)
diff(weight)
[1] 20 30 40

since 40 - 20 = 20, 70 - 40 = 30 and so on similarly try difftime for time series in case you need that

Upvotes: 1

Related Questions