Using Time Series in R

Question

I believe that using time-series in R has been discussed at length at Time series in R .

However, the dataset in above assumes a numeric array in all the SO posts and books I have read so far ( https://media.readthedocs.org/pdf/a-little-book-of-r-for-time-series/latest/a-little-book-of-r-for-time-series.pdf ). What if my data has categorical data as well? For instance,

> head(sassign)
  acctnum gender state   zip zip3 first last book_ nonbook_ total_ purch child youth cook do_it refernce
1   10001      M    NY 10605  106    49   29   109      248    357    10     3     2    2     0        1
2   10002      M    NY 10960  109    39   27    35      103    138     3     0     1    0     1        0
3   10003      F    PA 19146  191    19   15    25      147    172     2     0     0    2     0        0
4   10004      F    NJ 07016  070     7    7    15      257    272     1     0     0    0     0        1
5   10005      F    NY 10804  108    15   15    15      134    149     1     0     0    1     0        0
6   10006      F    NY 11366  113     7    7    15       98    113     1     0     1    0     0        0
  art geog buyer
1   0    2    no
2   0    1    no
3   0    0    no
4   0    0    no
5   0    0    no
6   0    0   yes

Now, here's what I did to create time-series object from above:--my objective is to group rows using "last" and then apply time-series type of object to "last" using sassign.

t_sassign <-data.frame(group_by(sassign,last))
t_sassign<-ts(t_sassign,start = c(2014,1),frequency = 12)

"Last" is the column indicating the last 'n' months since purchase. The above code works well except that the code is throwing warnings.

Warning message:
In data.matrix(data) : NAs introduced by coercion

Why is this happening? Please help me...My hypothesis is that I am getting NAs because R doesn't know how to group mixed data--grouping columns such as state (categorical) vs book_(continous). Am I correct?

However, if my hypothesis is correct, I am not quite sure how I can handle mixed data. Had it been all categorical, I would have used CrossTabs. Had it been all continous, I would have used functions such as sum, median etc. However, with mixed data, I am not quite sure.

I'd truly appreciate your thoughts.

Answer 1

No. "NA" is maybe because ts fails to convert character values of "gender", "state" and "buyer" to numeric. When they are factors, no warning message appear.

sassign = read.table(header = TRUE, text = "
acctnum gender state   zip zip3 first last book_ nonbook_ total_ purch child youth cook do_it refernce art geog buyer
1   10001      M    NY 10605  106    49   29   109      248    357    10     3     2    2     0        1 0    2    no
2   10002      M    NY 10960  109    39   27    35      103    138     3     0     1    0     1        0 0    1    no
3   10003      F    PA 19146  191    19   15    25      147    172     2     0     0    2     0        0 0    0    no
4   10004      F    NJ 07016  070     7    7    15      257    272     1     0     0    0     0        1 0    0    no
5   10005      F    NY 10804  108    15   15    15      134    149     1     0     0    1     0        0 0    0    no
6   10006      F    NY 11366  113     7    7    15       98    113     1     0     1    0     0        0 0    0   yes
");
t_sassign <-data.frame(group_by(sassign,last))
t_sassign<-ts(t_sassign,start = c(2014,1),frequency = 12)