[R] Tables package - remove NAs and NaN

[Duncan Murdoch]

On 13-04-22 10:40 PM, David Winsemius wrote:
>
> On Apr 22, 2013, at 5:49 PM, Santosh wrote:
>
>> Dear Rxperts,
>> q <- data.frame(p=rep(c("A","B"),each=10,len=30),
>> a=rep(c(1,2,3),each=10),id=seq(30),
>> b=round(runif(30,10,20)),
>> c=round(runif(30,40,70)))
>> The operation below...
>> tabular(((p=factor(p))*(a=factor(a))+1) ~ (N = 1) + (b + c)*
>> (mean+sd),data=q)
>> yields some rows of NAs and NaN as shown below
>>
>>              b               c
>> p a   N  mean  sd    mean  sd
>> A 1   10 16.30 2.497 52.30  9.358
>>    2    0   NaN    NA   NaN     NA
>>    3   10 15.60 2.716 60.30  8.001
>> B 1    0   NaN    NA   NaN     NA
>>    2   10 15.40 2.366 57.70 10.414
>>    3    0   NaN    NA   NaN     NA
>>    All 30 15.77 2.473 56.77  9.601
>>
>> How do I remove the rows having N=0 ?
>> I would like the resulting table look like..
>>              b               c
>> p a   N  mean  sd    mean  sd
>> A 1   10 16.30 2.497 52.30  9.358
>>      3   10 15.60 2.716 60.30  8.001
>> B  2   10 15.40 2.366 57.70 10.414
>>    All 30 15.77 2.473 56.77  9.601
>
> Here's a bit of a hack:
>
> tabular( (`p a`=interaction(p,a, drop=TRUE, sep=" ")) ~ (N = 1) + (b + c)*
>      (mean+sd),data=q)
>
>          b           c
>   p a N  mean sd     mean sd
>   A 1 10 12.8 0.7888 52.1 8.020
>   B 2 10 16.3 3.0569 54.9 8.711
>   A 3 10 14.6 3.7771 56.5 6.980
>
> I have been rather hoping that Duncan Murdoch would have noticed the earlier thread, but maybe he can comment on whether there is a more direct route/
>

This isn't something that the package is designed to handle:  if you say 
p*a, it wants all combinations of p and a.

If I wanted a table like that, I'd use a different hack.  One 
possibility is to create that interaction column, but display it as just 
the initial letter, labelled p, and then add another column to contain 
the a values as data.  It would be tricky to get the formatting right.

Another possibility is to generate the whole table with the N=0 rows, 
and then post-process it to remove those rows, and adjust the row labels 
appropriately.  This approach probably gives the nicer result, but the 
post-processing is quite messy:  you need to delete some rows from the 
table, from its rowLabels attribute, and from the justification 
attributes of both the table and its rowLabels.  (I should add a [ 
method to the package to hide this messiness.)

A third hack is less of a hack:  don't do that.  Just format the mean 
and sd in a way that displays the NA and NaN values as blanks.

Duncan Murdoch