Group (factorial) data with multiple factors. error: incompatible size (0), expecting 1 (the group size) or 1

Question

This post is a following up of Changing line color in ggplot based on "several factors" slope

I would like to group the data (bellow) by "PQ", however I get the following error:

"incompatible size (0), expecting 1 (the group size) or 1"

Data

ID<-c("A_P1","A_P1","A_P1","A_P1","A_P1","A_P2","A_P2","A_P2","A_P2","A_P2","A_P2","B_P1","B_P1","B_P1","B_P1","B_P1","B_P1","B_P1","B_P1","B_P2","B_P2","B_P2","B_P2","B_P2","B_P2","B_P2","B_P2")
Q<-c("C1","C1","C2","C3","C3","C1","C1","C2","C2","C3","C3","Q1","Q1","Q1","Q1","Q3","Q3","Q4","Q4","Q1","Q1","Q1","Q1","Q3","Q3","Q4","Q4")
PQ<-c("A_P1C1","A_P1C1","A_P1C2","A_P1C3","A_P1C3","A_P2C1","A_P2C1","A_P2C2","A_P2C2","A_P2C3","A_P2C3","B_P1Q1","B_P1Q1","B_P1Q1","B_P1Q1","B_P1Q3","B_P1Q3","B_P1Q4","B_P1Q4","B_P2Q1","B_P2Q1","B_P2Q1","B_P2Q1","B_P2Q3","B_P2Q3","B_P2Q4","B_P2Q4")
AS<-c("CF","CF","CF","CF","CF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF","CTF")
N<-c("N2","N3","N3","N2","N3","N2","N3","N2","N3","N2","N3","N0","N1","N2","N3","N1","N3","N0","N1","N0","N1","N2","N3","N1","N3","N0","N1")
Value<-c(4.7,8.61,8.34,5.89,8.36,1.76,2.4,5.01,2.12,1.88,3.01,2.4,7.28,4.34,5.39,11.61,10.14,3.02,9.45,8.8,7.4,6.93,8.44,7.37,7.81,6.74,8.5)

df<-data.frame(ID=ID,Q=Q,PQ=PQ,AS=AS,N=N,Value=Value)

Code that is delivering the error

#calculate slopes for N0 and N1
    df %>% 
      filter(N=="N0" | N=="N1") %>%
      group_by(PQ) %>%
      # use diff to calculate slope
      mutate(slope = diff(Value)) -> dat01

#calculate slopes for N0 and N2
    df %>% 
      filter(N=="N0" | N=="N2") %>%
      group_by(PQ) %>%
      # use diff to calculate slope
      mutate(slope = diff(Value)) -> dat02

Additionally, I would like to calculate the slope of the the remaining "PQ" factors (when existent), i.e. N0-N3;N1-N2 ... and so on

Answer 1

The error is due to the difference in length from the output of diff with respect to the original dataset. It returns one element less than the original data. So appending a 0 or NA will solve the issue

df %>% 
   filter(N=="N0" | N=="N1") %>%
   group_by(PQ) %>% 
   mutate(slope = c(0, diff(Value)))

To make it compact, instead of ==, we can use %in% when there are multiple elements

df %>%
   filter(N %in%  paste0("N", 0:1)) %>%
   group_by(PQ) %>%
   mutate(slope = c(0, diff(Value)))

---

Regarding the second issue, about doing this for all the combinations in 'N', use the combn on the unique elements of 'N', filter the 'N' based on the combination values, after grouping by 'PQ', calculate the diff of 'Value'. The output will be a list as we specified simplify = FALSE.

combn(as.character(unique(df$N)),2, FUN = function(x) df %>% 
            filter(N %in% x) %>% 
            group_by(PQ) %>%
            mutate(slope = c(0, diff(Value))), simplify = FALSE )