1

I have a dataframe of repetaed measures in WIDE format, with an ID variable (pid) some non time-varyant variables (i.e. age and sex) and mostly time-varyant variables, all named similarly with the following scheme: spec_var_0 spec_var_1 and spec_var_2 (we have 3 timewaves). Here it is an example:

structure(list(pid = structure(1:6, .Label = c("1", "2", "3", 
"4", "5", "6", "7", "8", "9", "10", "11", "12", "13", "14", "15", 
"16", "17", "18", "19", "20"), class = "factor"), dem_age = c(50, 
73, 55, 66, 65, 70), dem_sex = structure(c(1L, 2L, 1L, 2L, 2L, 
2L), .Label = c("Female", "Male"), class = "factor"), hx_steatosis = structure(c(1L, 
1L, 1L, 1L, 1L, 1L), .Label = c("No", "Yes"), class = "factor"), 
    bio_weight_0 = c(89.4, 73.9, 78.7, 86.2, 132, 90.5), bio_weight_1 = c(94.9, 
    66, 75.2, 82, 119, 81.6315789473684), bio_weight_2 = c(92, 
    56, 71, 81, 118, 86.5), lab_glyc_0 = c(5.52545293913178, 
    4.88280192258637, 4.79579054559674, 4.94875989037817, 5.17614973257383, 
    5.0998664278242), lab_glyc_1 = c(5.08140436498446, 5.01063529409626, 
    5.01727983681492, 4.77912349311153, 4.95582705760126, 5.03695260241363
    ), lab_glyc_2 = c(4.39444915467244, 5.06890420222023, 4.52178857704904, 
    4.62497281328427, 5.34710753071747, 4.95582705760126)), row.names = c(NA, 
6L), class = "data.frame")

Now I would like to convert to pivot_longer so to have the following format (sorry for the image, I do not know how to build such a table in R)

enter image description here

Searching Stack and other forums it seems that the issue could be fixed using refular expressions,but I have no idea of how to use them (in general and in this context specifically).

Any help would be (as usual) greatly appreciated

Anoushiravan R
  • 21,622
  • 3
  • 18
  • 41
Diego
  • 127
  • 2
  • 13

2 Answers2

2

I haven't quite taken you at your word as I have kept the number associated with the bio_weight in a column called bio_weight_num. But if you truly do not want it, you can just drop that column:

library(tidyr)
library(dplyr)

df  %>% 
    select(-c(hx_steatosis, starts_with("lab_glyc")))  %>% 
    pivot_longer(
        cols = starts_with("bio_weight"),
        names_to = "bio_weight_num", 
        values_to = "bio_weight"
    )   %>% 
    mutate(
        bio_weight_num = gsub("\\D", "", bio_weight_num)
    )

# A tibble: 18 x 5
#    pid   dem_age dem_sex bio_weight_num bio_weight
#    <fct>   <dbl> <fct>   <chr>               <dbl>
#  1 1          50 Female  0                    89.4
#  2 1          50 Female  1                    94.9
#  3 1          50 Female  2                    92
#  4 2          73 Male    0                    73.9
#  5 2          73 Male    1                    66  
#  6 2          73 Male    2                    56
#  7 3          55 Female  0                    78.7
#  8 3          55 Female  1                    75.2
#  9 3          55 Female  2                    71
# 10 4          66 Male    0                    86.2
# 11 4          66 Male    1                    82
# 12 4          66 Male    2                    81
# 13 5          65 Male    0                   132
# 14 5          65 Male    1                   119
# 15 5          65 Male    2                   118
# 16 6          70 Male    0                    90.5
# 17 6          70 Male    1                    81.6
# 18 6          70 Male    2                    86.5

I would suggest reading: https://tidyr.tidyverse.org/reference/pivot_longer.html

SamR
  • 8,826
  • 3
  • 11
  • 33
  • This is great ! Thank you for the superfast reply. Just last tweak (hoping is easy): I need to apply the same routine to a bunch of variables, (including among the other, lab_glyc), how can I dynamically change cols, names_to and values_to so to refer to different variable names (with the same structure,like bio_something_0, bio_something_1, bio_something_2 and lab_something_0, lab_something_1, lab_something_2) and have the values on bio_something, lab_something as you did for weight ? Again, thank you very much – Diego Mar 16 '22 at 09:21
  • This is a different question and the best way to do it depends on how many variable names you have, and their format (e.g. do they all end with a number?). It's going to be hard to answer it in a comment - have a think about whether you can extrapolate this code to your code requirements after reading the docs and if you can't then post it as a new question with the specifics of the data and what you want out of it. I would recommend looking at this: https://dcl-wrangle.stanford.edu/pivot-advanced.html#multiple-values – SamR Mar 16 '22 at 09:37
  • Thank you, will read docs you suggested and do some tests – Diego Mar 16 '22 at 09:46
1

Ok,

In this specific case (and not knowing Regular Expressions) I ended up renaming all variables in the dataframe to have a consistent but slightly different names structure: from spec_var_time to spec_var.time --> bio_weight_0 to bio_weight.0. This was useful in order to use names_sep option in pivot_longer. The following code works as expected:

testdf.l <- pivot_longer(data = testdf,
                         cols = c("bio_weight.0", "bio_weight.1", "bio_weight.2",
                                  "lab_glyc.0", "lab_glyc.1", "lab_glyc.2"), 
                         names_to = c('.value', 'timewave'),
                         names_sep = '\\.')

Again, thanks to SamR for the hints, and for having pointed me in the right direction

Diego
  • 127
  • 2
  • 13