使用分组的 dplyr 标识符在 2 个数据帧之间查找最匹配的数字

Question

I have 2 datasets each with a 'Patient ID` and a collection date measured from the same date "from start." In order to join these dataframes together, I'd like to match each sample in d1 to it's closest neighbor in d2. How can this be done with a function in dplyr?

d1<-data.frame(`Patient ID`=c(rep("001",4),rep("002",5)),`fromstart`=c(-5,30,90,150,-10,15,45,100,250),check.names = F)
d2<-data.frame(`Patient ID`=c(rep("001",7),rep("002",4)),`fromstart`=c(-20,10,30,50,90,110,150,-10,15,45,100),check.names = F)

closest_date<-function(cases,d2) {
  return(d2 %>% select(`Patient ID`,fromstart) %>% unique() %>% filter(`Patient ID`==cases$`Patient ID`) %>% rowwise() %>% mutate(date_match=as.numeric(cases$fromstart[which.min(abs(fromstart - cases$fromstart))])))
}

d1 %>% select(`Patient ID`,fromstart) %>% unique() %>% group_by(`Patient ID`) %>% rowwise() %>% mutate(closest=closest_date(.,d2))

Answer 1

If I understood your problem correctly you want to join by patient ID and then select those lines where the difference between fromstart is the smallest? If so this would be a solution

library(dplyr)
d1 %>% 
  dplyr::full_join(d2, by = c("Patient ID"), suffix = c("_1", "_2")) %>% 
  dplyr::mutate(DIF = abs(fromstart_1  - fromstart_2)) %>% 
  dplyr::group_by(`Patient ID`, fromstart_1) %>% 
  dplyr::filter(DIF == min(DIF))

As you can see this does not really work well if you want unique combinations because there can be cases where the distance is the same... than again maybe I did not get your questoin right

Instead of using the absolute value you could filter for positive differences/distances as well if you want fromstart of the second table to be larger than from the first, this would reduce double entries to a certain degree