rdplyrreshape

Keep all matched rows when reshaping from long to wide

I am trying to reshape my data from long to wide format. I would like to keep all rows that match for sn3=1 and sn3=2 instead of taking only the first value. How can I achieve this?

Desired output:

sn2 leisure.1 hmonth.1 hyear.1 leisure.2 hmonth.2 hyear.2
227 230 6 2000 540 6 2000
227 130 6 2000 170 6 2000
250 370 6 2000 380 6 2000
250 380 6 2000 190 6 2000
294 120 6 2000 210 6 2000
294 200 6 2000 310 6 2000
307 130 7 2000 220 7 2000
307 480 7 2000 270 7 2000 
> dput(df)
structure(list(sn2 = structure(c(227, 227, 227, 227, 249, 249, 
250, 250, 250, 250, 294, 294, 294, 294, 307, 307, 307, 307), label = "household number", format.stata = "%8.0g"), 
    sn3 = structure(c(1, 1, 2, 2, 1, 1, 1, 1, 2, 2, 1, 1, 2, 
    2, 1, 1, 2, 2), label = "person number", format.stata = "%8.0g"), 
    leisure = c(230, 130, 540, 170, 430, 480, 370, 380, 380, 
    190, 120, 200, 210, 310, 130, 480, 220, 270), hmonth = structure(c(6, 
    6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7), label = "month of household interview", format.stata = "%8.0g", labels = c(january = 1, 
    february = 2, march = 3, april = 4, may = 5, june = 6, july = 7, 
    august = 8, september = 9, october = 10, november = 11, december = 12
    ), class = c("haven_labelled", "vctrs_vctr", "double")), 
    hyear = structure(c(2000, 2000, 2000, 2000, 2000, 2000, 2000, 
    2000, 2000, 2000, 2000, 2000, 2000, 2000, 2000, 2000, 2000, 
    2000), label = "year of household interview", format.stata = "%8.0g")), row.names = c(NA, 
-18L), class = c("tbl_df", "tbl", "data.frame"), na.action = structure(c(`3492` = 3492L, 
`3493` = 3493L, `3494` = 3494L, `3495` = 3495L, `3496` = 3496L, 
`3497` = 3497L, `3498` = 3498L, `3499` = 3499L, `3500` = 3500L, 
`3501` = 3501L, `3508` = 3508L, `3509` = 3509L, `3510` = 3510L, 
`3511` = 3511L, `3512` = 3512L, `3513` = 3513L, `3518` = 3518L, 
`3519` = 3519L, `3520` = 3520L, `3521` = 3521L, `3522` = 3522L, 
`3523` = 3523L, `3524` = 3524L, `3525` = 3525L), class = "omit"))
Solution

  • As long as you get rid of the rows for which each sn2 only has a single value of sn3, you can just pivot_wider then unnest_longer:

    library(tidyverse)

df %>%
  filter(n() > 2L, .by = sn2) %>%
  pivot_wider(names_from = sn3, values_from = leisure:hyear, 
              values_fn = list) %>%
  unnest_longer(leisure_1:hyear_2) 
#> # A tibble: 8 x 7
#>     sn2 leisure_1 leisure_2  hmonth_1  hmonth_2 hyear_1 hyear_2
#>   <dbl>     <dbl>     <dbl> <dbl+lbl> <dbl+lbl>   <dbl>   <dbl>
#> 1   227       230       540  6 [june]  6 [june]    2000    2000
#> 2   227       130       170  6 [june]  6 [june]    2000    2000
#> 3   250       370       380  6 [june]  6 [june]    2000    2000
#> 4   250       380       190  6 [june]  6 [june]    2000    2000
#> 5   294       120       210  6 [june]  6 [june]    2000    2000
#> 6   294       200       310  6 [june]  6 [june]    2000    2000
#> 7   307       130       220  7 [july]  7 [july]    2000    2000
#> 8   307       480       270  7 [july]  7 [july]    2000    2000