zoo column name for single column object

Question

I have a question on column names in zoo. I usually create zoo objects from a data frame, and I pick up the column(s) from the data frame to be the zoo column(s). What I found is that, if I only specify one column to the zoo object, then the column name will not be taken by zoo. Does that mean it is not considered a "column" in zoo?

Here is an example how I usually do it, with one and two columns.

Lines.1 = "Index,dbt
2008-08-20 15:03:18,88.74
2008-08-20 15:08:18,88.74
2008-08-20 15:13:18,86.56
2008-08-20 15:18:18,85.82"

Lines.2 = "Index,dbt,rh
2008-08-20 15:03:18,88.74,18.25
2008-08-20 15:08:18,88.74,17.25
2008-08-20 15:13:18,86.56,18.75
2008-08-20 15:18:18,85.82,19.75"

x =read.table(text = Lines.1, header = TRUE, sep = ",")
y =read.table(text = Lines.2, header = TRUE, sep = ",")

colnames(x)
colnames(y)

library(zoo)
zx = zoo(x[,2], as.POSIXct(x$Index, tz="GMT"))
zy = zoo(y[,2:3], as.POSIXct(y$Index, tz="GMT"))

colnames(zx)
colnames(zy)

The result shows as follows:

> colnames(zx)
NULL
> colnames(zy)
[1] "dbt" "rh" 

Do I miss something?

Answer 1

This is the default behaviour of [ when used with arrays or data frames; empty dimensions are dropped. Consider

> x[, 2]
[1] 88.74 88.74 86.56 85.82
> class(x[,2])
[1] "numeric"
> is.data.frame(x[,2])
[1] FALSE

In this case the 1-column data frame doesn't need information about which column it is and hence R drops that information and returns the contents of the column as a numeric (in this case) vector, as can be seen above. That vector doesn't have a colname attribute and hence zoo has nothing to work with.

A solution is to use drop = FALSE in the index x[, 2, drop = FALSE] as in

> zx <- zoo(x[, 2, drop = FALSE], as.POSIXct(x$Index, tz="GMT"))
> zx
                      dbt
2008-08-20 15:03:18 88.74
2008-08-20 15:08:18 88.74
2008-08-20 15:13:18 86.56
2008-08-20 15:18:18 85.82

To see why/how this works, look at

> x[, 2, drop = FALSE]
    dbt
1 88.74
2 88.74
3 86.56
4 85.82
> is.data.frame(x[, 2, drop = FALSE])
[1] TRUE

And note the lack of colnames when the default (TRUE) is used in the [ index:

> colnames(x[, 2, drop = FALSE])
[1] "dbt"
> colnames(x[, 2, drop = TRUE])
NULL

Now read ?'[' for more details.

Answer 2

This behavior is not due to zoo. x[,2] does not return a data frame but a vector. Hence, there are no columns and no column names.

x[,2]
[1] 88.74 88.74 86.56 85.82

If you want to return a one-column data frame you can use x[2] without a comma or x[,2, drop = FALSE].

x[2]
    dbt
1 88.74
2 88.74
3 86.56
4 85.82

x[,2, drop = FALSE]
    dbt
1 88.74
2 88.74
3 86.56
4 85.82

The default value of drop is TRUE. This means that a one-column data frame is automatically transformed into a vector.

Now, it works:

zx <- zoo(x[2], as.POSIXct(x$Index, tz="GMT"))
colnames(zx)
[1] "dbt"