R长到宽格式的分类变量和日期

Question

I want to make a dataset with long format with a categorial variabel (medication) that has a start and end date to wide format. As the result there should be one line per ID and for each medication a column that has the entry 1/0 either the patient got the medication or not. And the medication column should have the respective start and end date as extra columns.

I wanted to

test <- data.frame(
  PatID  = c(1L, 1L, 2L, 2L, 3L, 4L,4L),
  medication = c("Jak","Others", "HU", "Inf","Others", "HU","Others"),  startDate   = c("2016-12-14", "2017-02-04", "2016-03-26", "2016-06-13", "2012-27-03", "2012-04-21", "2010-02-03"),
  endDate   = c("2018-11-14", "2018-02-25", "2017-06-13", "2017-11-12", "2018-27-03", "2016-04-30", "2016-08-16")
)

The output should be the following

ID   Jak   Jak_startDate   Jak_endDate   HU   HU_startDate   HU_endDate   Inf   Inf_startDate   Inf_endDate   Others   Others_startDate   Others_endDate <br/>
1    1      2016-12-14    2018-11-14     0     NA                NA        0        NA             NA            1      2017-02-04        2018-02-25
2    0         NA               NA       1   2017-06-13   2017-11-12       1   2018-03-27   2016-04-30           0         NA               NA
3    0         NA               NA       0     NA                NA        0        NA             NA            1      2012-27-03        2018-27-03
4    0         NA               NA       1   2012-04-21   2016-04-30       0        NA             NA            1      2010-02-03        2016-08-16

Answer 1

Using tidyverse here is what I did:

test %>% 
gather(key, value, - PatID, -medication) %>% 
arrange(PatID, value) %>% 
mutate(new_key = paste(medication, key, sep = "_")) %>% 
select(PatID, new_key, value) %>% 
spread(new_key, value) %>% 
left_join(test %>% select(PatID, medication) %>% 
mutate(ind = 1) %>% 
spread(medication, ind))

Here, I make the data longer, then arrange it according the PatID and value . Then I create a new key column new_key and select only three variables: PatID , new_key , and value . Then I turn this all into wide data, but we still need the columns Jak , HU , etc. which seem to be indicator variables. For this within the left_join I take the test data and spread that as well to get the columns you have requested.