Odtnhj

I could really use some help with the following:

I some sample data:

Group<-c("A","A","B","B","C","C","D", "D")
Value1<-c("7","1",8,7,"NA",9,10,12)
Value2<-c("NA","NA","NA",7,3,9,7,4)
df<-data.frame(Group, Value1, Value2)

 Group Value1 Value2
 A 7 NA
 A 1 NA
 B 8 NA
 B 7 7
 C NA 3
 C 9 9
 D 10 7
 D 12 4

I want delete all groups that are either fully filled in for Value1 and Value2 (like Group D) or have no data for Value2 within a group (like Group A). So that I end up with the groups that are filled in for both Value1 and Value2 but are incomplete, like:

 Group Value1 Value2
 B 8 NA
 B 7 7
 C NA 3
 C 9 9

I Know how to delete NA's, and I have tried some things with

setDT()[, := if(any(Value2==)) "" else "" , by = .()]

but I don't really now how to use it in this case..

Does someone know how I could do this?

Something like:

setDT(df)[, .SD[!all(is.na(Value2)) & (anyNA(Value1) | anyNA(Value2))], Group]

Output:

 Group Value1 Value2
1: B 8 NA
2: B 7 7
3: C <NA> 3
4: C 9 9

Note that in order for this to work, your values should indeed be NA and not just "NA" as strings, i.e. it works with a modified dataframe example:

Group<-c("A","A","B","B","C","C","D", "D")
Value1<-c("7","1",8,7,NA,9,10,12)
Value2<-c(NA,NA,NA,7,3,9,7,4)
df<-data.frame(Group, Value1, Value2)

Using data from @arg0naut (to keep NA as NA and not "NA") a base R solution using two ave would be

df[!with(df, ave(complete.cases(df), Group, FUN = all) | 
 ave(is.na(Value2), Group, FUN = all)), ]

# Group Value1 Value2
#3 B 8 NA
#4 B 7 7
#5 C NA 3
#6 C 9 9

We keep two conditions separately and find the rows which we don't want to select and then take negation of it.