将日期转换为时间序列的月/年格式

Question

I have some have some water quality sample data.

> dput(GrowingArealog90s[1:10,])
structure(list(SampleDate = structure(c(6948, 6949, 6950, 7516, 
7517, 7782, 7783, 7784, 8092, 8106), class = "Date"), Flog90 =  c(1.51851393987789, 
1.48970743802793, 1.81243963000062, 0.273575501327576, 0.874218895695207, 
1.89762709129044, 1.44012088794774, 0.301029995663981, 1.23603370361931, 
0.301029995663981)), .Names = c("SampleDate", "Flog90"), class = c("tbl_df", 
"data.frame"), row.names = c(NA, -10L))

This data is collected monthly, although some months are missed over the 25 year period.

I know there is so much help out there for converting dates to different formats but I have not been able to figure this out. I want to create a time series with just a month/year format, so that I can do things like decompose the data by month and run seasonal kendalls and such. I have tried so many different ways of converting my date to the desired format that I have completely confused myself. I don't care about the exact format as long as it is recognized month/year.

I also need to fill in the missing months with NAs.

I tried uploading the "SampleDate" column in a numeric format, "yyyymm". I could then merge that data frame with another that contained all the dates I need.

GA90 <- merge(Dates, GrowingArealog90s, by.x = "Date", by.y = "Date", all.x = TRUE)

However, when I converted the resulting data frame to a time series it would not recognize the 12 month frequency.

 GA90ts <- as.ts(GA90, frequency(12))

> GA90ts
Time Series:
Start = 1 
End = 324 
Frequency = 1 

Any help with this is appreciated.

Answer 1

Here's how to do it with zoo . You'll get a warning, but it's OK for now. You'll get a series with mon/yy.

series <-structure(list(SampleDate = structure(c(6948, 6949, 6950, 7516,
7517, 7782, 7783, 7784, 8092, 8106), class = "Date"), Flog90 =  c(1.51851393987789,
1.48970743802793, 1.81243963000062, 0.273575501327576, 0.874218895695207,
1.89762709129044, 1.44012088794774, 0.301029995663981, 1.23603370361931,
0.301029995663981)), .Names = c("SampleDate", "Flog90"), class = c("tbl_df",
"data.frame"), row.names = c(NA, -10L))

library(zoo)
series <-as.data.frame(series) #to drop dplyr class
series.zoo <-zoo(series[,-1,drop=FALSE],as.yearmon(series[,1]))

Best practice would be to keep your series with actual date and use as.yearmon or as.yearmon only when you actually need to make calculations or aggregate.zoo by month and year.

The following is a matter of taste, but I've dealt with a lot of time series and I think zoo is superior to ts and xts . Much more flexible.

Now, to fill in missing values, you have to create a vector of dates. Here, I'm using a zoo object with actual dates. I then use na.locf , which is "last observation carry forward". You could also look at na.approx .

series.zoo <-zoo(series[,-1,drop=FALSE],(series[,1]))
my.seq <-seq.Date(first(series[,1,drop=FALSE]), last(series[,1,drop=FALSE]),by="month")
merged <-merge.zoo(series.zoo,zoo(,my.seq))
na.locf(merged)

UPDATE

With aggregate.

GrowingArealog90s <-structure(list(SampleDate = structure(c(6948, 6949, 6950, 7516,
7517, 7782, 7783, 7784, 8092, 8106), class = "Date"), Flog90 =  c(1.51851393987789,
1.48970743802793, 1.81243963000062, 0.273575501327576, 0.874218895695207,
1.89762709129044, 1.44012088794774, 0.301029995663981, 1.23603370361931,
0.301029995663981)), .Names = c("SampleDate", "Flog90"), class = c("tbl_df",
"data.frame"), row.names = c(NA, -10L))

library(zoo);library(xts)
GrowingArealog90s <-as.data.frame(GrowingArealog90s) #to remove dplyr format
GrowingArealog90s.zoo <-zoo(GrowingArealog90s[,-1,drop=FALSE],as.Date(GrowingArealog90s[,1]))

#First aggregate by month. I chose to get the mean per month
GrowingArealog90s.agg <-aggregate(GrowingArealog90s.zoo, as.yearmon, mean) #replace mean with last to get last reading of the month

#Then create a sequence of months and merge it
my.seq <-seq.Date(first(GrowingArealog90s[,1]), last(GrowingArealog90s[,1]),by="month")
merged <-merge.zoo(GrowingArealog90s.agg ,zoo(,as.yearmon(my.seq)))
na.locf(merged)