I am trying to find a vectorized, general way to accomplish a task in R for which I typically use Stata. I have a list of variables which have a common structure, and I am hoping to apply the same transformations to all of them efficiently. There are two versions of each variable (x and m), each of which only have a value (some numeric, some char) if the observation is in group 1 or 2, respectively. I am trying to create a general variable which contains only the corresponding non-NA value for the row. In Stata, I would accomplish this using the below code:
foreach word in cost cost_add25 cost_less25 parkingaccess parkingaccess_add25 parkingaccess_less25 timeresult timeresult_add25 timeresult_less25 timetravel timetravel_add25 timetravel_less25 timewait timewait_add25 timewait_less25 freeparking onlinesch service service_less1 service_plus1 pcprec specialty {
gen statusquo_`word' = statusquo_x_`word' if group == 1
replace statusquo_`word' = statusquo_m_`word' if group == 2
}
However I have so far been stumped on how to efficiently do this in R with a vectorized function-- I have tried using purrr:map, lapply, and mapply, but have come up short. Here is my current implementation, which is pretty gross:
database <- database %>%
mutate(
statusquo_cost = case_when(group == 1 ~ statusquo_x_cost , group == 2 ~ statusquo_m_cost ),
statusquo_cost_add25 = case_when(group == 1 ~ statusquo_x_cost_add25 , group == 2 ~ statusquo_m_cost_add25 ),
statusquo_cost_less25 = case_when(group == 1 ~ statusquo_x_cost_less25 , group == 2 ~ statusquo_m_cost_less25 ),
statusquo_parkingaccess = case_when(group == 1 ~ statusquo_x_parkingaccess , group == 2 ~ statusquo_m_parkingaccess ),
statusquo_parkingaccess_add25 = case_when(group == 1 ~ statusquo_x_parkingaccess_add25 , group == 2 ~ statusquo_m_parkingaccess_add25 ),
statusquo_parkingaccess_less25 = case_when(group == 1 ~ statusquo_x_parkingaccess_less25 , group == 2 ~ statusquo_m_parkingaccess_less25 ),
statusquo_timeresult = case_when(group == 1 ~ statusquo_x_timeresult , group == 2 ~ statusquo_m_timeresult ),
statusquo_timeresult_add25 = case_when(group == 1 ~ statusquo_x_timeresult_add25 , group == 2 ~ statusquo_m_timeresult_add25 ),
statusquo_timeresult_less25 = case_when(group == 1 ~ statusquo_x_timeresult_less25 , group == 2 ~ statusquo_m_timeresult_less25 ),
statusquo_timetravel = case_when(group == 1 ~ statusquo_x_timetravel , group == 2 ~ statusquo_m_timetravel ),
statusquo_timetravel_add25 = case_when(group == 1 ~ statusquo_x_timetravel_add25 , group == 2 ~ statusquo_m_timetravel_add25 ),
statusquo_timetravel_less25 = case_when(group == 1 ~ statusquo_x_timetravel_less25 , group == 2 ~ statusquo_m_timetravel_less25 ),
statusquo_timewait = case_when(group == 1 ~ statusquo_x_timewait , group == 2 ~ statusquo_m_timewait ),
statusquo_timewait_add25 = case_when(group == 1 ~ statusquo_x_timewait_add25 , group == 2 ~ statusquo_m_timewait_add25 ),
statusquo_timewait_less25 = case_when(group == 1 ~ statusquo_x_timewait_less25 , group == 2 ~ statusquo_m_timewait_less25 ),
statusquo_freeparking = case_when(group == 1 ~ statusquo_x_freeparking , group == 2 ~ statusquo_m_freeparking ),
statusquo_onlinesch = case_when(group == 1 ~ statusquo_x_onlinesch , group == 2 ~ statusquo_m_onlinesch ),
statusquo_service = case_when(group == 1 ~ statusquo_x_service , group == 2 ~ statusquo_m_service ),
statusquo_service_less1 = case_when(group == 1 ~ statusquo_x_service_less1 , group == 2 ~ statusquo_m_service_less1 ),
statusquo_service_plus1 = case_when(group == 1 ~ statusquo_x_service_plus1 , group == 2 ~ statusquo_m_service_plus1 ),
statusquo_pcprec = case_when(group == 1 ~ statusquo_x_pcprec , group == 2 ~ statusquo_m_pcprec ),
statusquo_specialty = case_when(group == 1 ~ statusquo_x_specialty , group == 2 ~ statusquo_m_specialty )
)
In an ideal world, I would have something like this, where "VAR" is cycled through the values in the above list, however R is very type-sensitive and treats the new variable name as text.
vars <- c("cost","cost_add25","cost_less25","parkingaccess","parkingaccess_add25","parkingaccess_less25","timeresult","timeresult_add25","timeresult_less25","timetravel","timetravel_add25","timetravel_less25","timewait","timewait_add25","timewait_less25","freeparking","onlinesch","service","service_less1","service_plus1","pcprec","specialty")
database <- database %>%
mutate(
statusquo_VAR = case_when(group == 1 ~ statusquo_x_VAR, group == 2 ~ statusquo_m_VAR)
)
Any suggestions would be very helpful! I can provide a data sample if needed, though I am not sure this question is complicated enough to need one.
I tried to use a vectorized operation to combine a list of two variables together, based on a structured list of their names, and am unable to do so without a cumbersome approach.
CodePudding user response：
I have modified your code, hopefully it will get what you wanted. Another minor point, I used ifelse to replace case_when, either should work though.
vars <- c("cost","cost_add25","cost_less25","parkingaccess","parkingaccess_add25","parkingaccess_less25","timeresult","timeresult_add25","timeresult_less25","timetravel","timetravel_add25","timetravel_less25","timewait","timewait_add25","timewait_less25","freeparking","onlinesch","service","service_less1","service_plus1","pcprec","specialty")
for (i in seq_along(vars)){
database <- database %>%
mutate(
!!paste0("statusquo_", vars[i]) := ifelse(group == 1, get(paste0("statusquo_x_", vars[i])), get(paste0("statusquo_m_", vars[i])))
)
}
Edit:
With newer version of dplyr >= 1.0
for (i in vars){
database <- database %>%
mutate(
"statusquo_{i}" := ifelse(group == 1, get(paste0("statusquo_x_", i)), get(paste0("statusquo_m_", i)))
)
}