如何提取单变量时间序列的时间属性？

Question

How can I convert a simple time series object to be of dimensions 2 by adding time attributes so that colnames can be used?

Edit :

In this question , the way to refer to time series was to use colnames and suggested method was to convert it to zoo objects. In case of single time series, if I do names(tseries), it gives NULL. So to assign the names, tried using colnames which can only be used in objects of dimension 2 and constructed by adding index of dates. Here I am trying to use the time attributes already there in the list object so that I dont have to convert back and forth from zoo objects.

If there is an alternative I will be very happy to know that. The purpose is to be able to use the name of the time series in title of plots etc.

Edit 2:

To rephrase the question, how to give name to a time series ? Using apply strips the time series attributes as asked in many questions here. Then to get back the time attributes and the name, and in a format which allows using colnames, what are the methods ? I started with series and a column of dates which allowed naming the series but in a vector format returned from the lapply(tslist, function), I need to reallocate the names. If there is a simpler way to do this, please do let me know as that is the main problem.

A somewhat related question is link where answer by @Henrik suggests adding a dummy column to avoid coercion of zoo object to a vector to keep the it of dimension 2.

In a specific situation where we have a list object with required time info, I followed this method.

When using HP filter to find the cycle, I do the following

x <-ts(rnorm(100), start = c(1990, 1), frequency = 12)

library(mFilter)
hp <- hpfilter(x, 1600)
c1 <- hp$cycle

str(c1)
 #Time-Series [1:100] from 1990 to 1998: 1.852 -0.368 -0.942 -0.756 1.006 ...

str(hp) # shows that it has the attributes of time series and tsp etc. 

Str of main object from which subsetting is

List of 10
 $ cycle  : Time-Series [1:100] from 1990 to 1998: 1.852 -0.368 -0.942 -0.756 1.006 ...
 $ trend  : ts [1:100, 1] -0.191 -0.193 -0.193 -0.191 -0.187 ...
  ..- attr(*, "dimnames")=List of 2
  .. ..$ : NULL
  .. ..$ : chr "Series 1"
  ..- attr(*, "tsp")= num [1:3] 1990 1998 12
 $ fmatrix: num [1:100, 1:100] 0.799 -0.178 -0.156 -0.135 -0.116 ...
 $ title  : chr "Hodrick-Prescott Filter"
 $ xname  : chr "x"
 $ call   : language hpfilter(x = x, freq = 1600)
 $ type   : chr "lambda"
 $ lambda : num 1600
 $ method : chr "hpfilter"
 $ x      : ts [1:100, 1] 1.661 -0.561 -1.135 -0.947 0.819 ...
  ..- attr(*, "dimnames")=List of 2
  .. ..$ : NULL
  .. ..$ : chr "Series 1"
  ..- attr(*, "tsp")= num [1:3] 1990 1998 12
 - attr(*, "class")= chr "mFilter"

c1 is an object of dimension 1 because when I try to assign it a name,

colnames(c1) <- "x" 

gives this error

Error in `colnames<-`(`*tmp*`, value = "iip") : 
  attempt to set 'colnames' on an object with less than two dimensions

To help use name of time series, we can either convert c1 to a zoo object by adding an index or cbind it with a column of dates.

Since the name and time attributes information is there in the starting object from which we are subsetting, is there a way to use it to extract a component with the time attributes ?

Edit :

The extracted series must have structure like that of inputted series which is there in the list object

ts [1:100, 1] 1.661 -0.561 -1.135 -0.947 0.819 ...
  ..- attr(*, "dimnames")=List of 2
  .. ..$ : NULL
  .. ..$ : chr "Series 1"
  ..- attr(*, "tsp")= num [1:3] 1990 1998 12

which I think can be accessed as

attributes(c1) <- attributes(hp$x)

colnames(c1) <- "X1"  # is now possible. 

How can I return hp$cycle along with its attributes ?

How can I access chr "Series 1" when allocating colnames ?

colnames(c1) 

# Series 1

Edit :

## Return hp$cycle along with the time attributes so that str(h1) is as follows

h1 <- hp$cycle
str(h1)
# this gives 
Time-Series [1:100] from 1990 to 1998: 1.852 -0.368 -0.942 -0.756 1.006 ..
#I want it to be of this form 

ts [1:100, 1] 1.661 -0.561 -1.135 -0.947 0.819 ...
      ..- attr(*, "dimnames")=List of 2
      .. ..$ : NULL
      .. ..$ : chr "Series 1"
      ..- attr(*, "tsp")= num [1:3] 1990 1998 12

Thanks a lot.

Edit :

These are the alternatives I used to extract the time attributes from one series and add to another series.

1. Using tsp : This was giving error or dimension mismatch

   tsp(h1) <- tsp(x)

2. Using start, end, frequency information in tsp Since tsp for monthly data is in fractions, didnt know how to get the starting month info ie converting 1970.250 to 1970, 4 say, (dont know which is the starting month) Then I could do something like

   h1.ts <- ts(h1, start =c(tsp 1 , n), frequency = tsp[3])

Answer 1

Based on the info provided, I guess you can try:

 h1 <- hp$cycle
 str(h1)
 #Time-Series [1:100] from 1990 to 1998: -0.516 1.101 -0.756 0.786 0.926 ...
 attributes(h1)
 #$tsp
 #[1] 1990.00 1998.25   12.00

 #$class
 #[1] "ts"

If you look at the str(hp) , it is a list of 10 elements. To access the attributes of different elements, you can use attributes or attr . For example:

 attr(hp$cycle, "tsp")
 #[1] 1990.00 1998.25   12.00

Or

  attributes(hp$cycle)[["tsp"]] #
 #[1] 1990.00 1998.25   12.00

To get the whole attributes for the list element

 attributes(hp$cycle)
 #$tsp
 #[1] 1990.00 1998.25   12.00

 #$class
 #[1] "ts"

Using c concatenates the attributes from different list elements and keep the list structure

 attributes(h1) <- c(attributes(hp$cycle), attributes(hp$trend))
 str(h1)
 #ts [1:100, 1] -0.516 1.101 -0.756 0.786 0.926 ...
 #- attr(*, "tsp")= num [1:3] 1990 1998 12
 #- attr(*, "dimnames")=List of 2
 # ..$ : NULL
 #..$ : chr "Series 1"

 tsp(h1)
 #[1] 1990.00 1998.25   12.00
 tsp(h1) <- tsp(x) #no errors here

Answer 2

You can use start , end and time :

  start(c1)
 [1] 1990    1

  end(c1)
 [1] 1998    4

  time(c1)
          Jan      Feb      Mar      Apr      May      Jun      Jul      Aug
1990 1990.000 1990.083 1990.167 1990.250 1990.333 1990.417 1990.500 1990.583
1991 1991.000 1991.083 1991.167 1991.250 1991.333 1991.417 1991.500 1991.583
1992 1992.000 1992.083 1992.167 1992.250 1992.333 1992.417 1992.500 1992.583
1993 1993.000 1993.083 1993.167 1993.250 1993.333 1993.417 1993.500 1993.583
1994 1994.000 1994.083 1994.167 1994.250 1994.333 1994.417 1994.500 1994.583
1995 1995.000 1995.083 1995.167 1995.250 1995.333 1995.417 1995.500 1995.583
1996 1996.000 1996.083 1996.167 1996.250 1996.333 1996.417 1996.500 1996.583
1997 1997.000 1997.083 1997.167 1997.250 1997.333 1997.417 1997.500 1997.583
1998 1998.000 1998.083 1998.167 1998.250                                    
          Sep      Oct      Nov      Dec
1990 1990.667 1990.750 1990.833 1990.917
1991 1991.667 1991.750 1991.833 1991.917
1992 1992.667 1992.750 1992.833 1992.917
1993 1993.667 1993.750 1993.833 1993.917
1994 1994.667 1994.750 1994.833 1994.917
1995 1995.667 1995.750 1995.833 1995.917
1996 1996.667 1996.750 1996.833 1996.917
1997 1997.667 1997.750 1997.833 1997.917
1998                                    

Answer 3

If the question is how to create an nx 1 monthly ts series with a name then:

xx <- ts(cbind(A = 1:5), start = c(2000, 1), freq = 12)

We can now refer to it by name:

xx[, "A"]

We can extract various components like this:

xx[2] # second point
xx[2, 1] # second point with name
xx[2, "A"] # same

start(xx) # start time
time(xx)[1]

end(xx) # end time
time(xx)[NROW(xx)]

time(xx) # times
c(time(xx))

frequency(xx)

cycle(xx) # months as a ts series
c(cycle(xx)) # plain

floor(time(xx)) # years as a ts series
c(floor(time(xx)))

We can convert yy below to an nx 1 series with a name:

yy <- ts(1:5, start = c(2000, 1), frequency = 12)

dim(yy) <- c(length(yy), 1)
colnames(yy) <- "A"

yy[2, 1] # get second point with name

Added If we have:

xx <- ts(cbind(A = 1:5), start = c(2000, 1), freq = 12)

and attributes get stripped:

xx2 <- c(xx)

then we can do this:

xx[] <- xx2

Answer 4

I am not sure what the desired output. Hope this helps. Please show the desired output.

newdf<-as.data.frame(hp$x)
colnames(newdf)
[1] "Series 1"