I've been importing data from a csv, then putting it into a dataframe. I then pad it, convert it to an XTS. After that I use apply.weekly to break it down into weekly data. The code is below.
library(padr)
library(xts)
library(tidyr)
library(seasonal)
library(forecast)]
library(lubridate)
library(tidyverse)
library(zoo)
df <- data.frame(data$f1.date[1:831], data.$f1[1:831]) #create data frame
colnames(df)<- c('date', 'f1')
df$date <- (as.Date(df$date, format='%m/%d/%Y'))
df<-pad(df) # add missing dates
df<-xts(df, order.by = df$date)
df$t1<-replace_na(df$f1,0) # Add Zeros for NA vals
dtw<-apply.weekly(df$f1,FUN=colSums) # I've tried using sum, colSums and mean. Mean works fine.
However, sum or colSums returns the following.
Error in FUN(x[(INDEX[y] + 1):INDEX[y + 1]], ...) : 'x' must be numeric
So far, my only thought was to try df$t1<-as.numeric(df$t1), but that had no effect.
Does anyone have any suggestions on this? This exact script was working fine with the same csv file, and I don't think anything has changed. All I can think is that some package didn't get loaded.
A sample of my data is below, the full version has approximately 800 rows. I am working on making it reproducible now.
Using dput(head(df)) gives me this.
structure(c("2010-01-01", "2010-01-02", "2010-01-03", "2010-01-04",
"2010-01-05", "2010-01-06", " 1", "0", "0", "0", "0", "0"), class = c("xts",
"zoo"), index = structure(c(1262304000, 1262390400, 1262476800,
1262563200, 1262649600, 1262736000), tzone = "UTC", tclass = "Date"), .Dim = c(6L,
2L), .Dimnames = list(NULL, c("date", "f1")))
I also edited it to change 't1' to 'f1' for consistency.
Thanks
The problem is that your column f1 is of class character.
is.character(df$f1)
[1] TRUE
One approach would be to convert it to numeric within your apply.weekly call:
apply.weekly(df$f1,function(x){sum(as.numeric(x))})
f1
2010-01-03 1
2010-01-06 0
The reason for all this is because xts class objects hold the data inside a matrix. And matrices in R can only have columns of one class.