Create case-control match by distance for conditional logistic regression in R

Question

Aloha,

I am planning to run a case-control study for study sites that are evenly distributed spatially around the country. I need to select each case in the dataset and then match it to x number of controls (we will use a sensitivity analysis to select the optimal matches, so I need to be able to run it for 1,2,3,4,5,6,7,8, etc number of controls). As there is a spatial element to the data I want to run this computation within a distance matrix by selecting the controls within 25000 meters of the case.

I cannot find the optimal algorithm to run this computation in R. Is anyone aware of an optimal R package that would help me achieve this?

Thank you

Answer 1

To solve this I did the following

Got the coordinates of the site centroid (x,y)

Split the DB into my case-control groups

ran a spatial buffer of the cases

ran an intersection of the controls

assigned a label to all intersections (match_no)

Randomly sampled from within the match_no column

Code below.

db1 <- read.csv("db1_clf.csv")

library(sf)
dat <- st_as_sf(x=db1,
                   coords = c("x_coor_farm", "y_coor_farm"),
                   crs= "+proj=longlat +datum=WGS84 +ellps=WGS84 +towgs84=0,0,0")


##Filter the positive cases
library(dplyr)
case = dat %>% filter(TB2017 == "1")
control = dat %>% filter(TB2017 == "0")

case_buff = st_buffer(case, dist = 25000)

case_int = st_intersection(case_buff, control)

library(dplyr)

case_int$match_no <- as.integer(factor(case_int$idunique))

library(dplyr)

pos_db <- case_int %>%
  select("idunique", "match_no")

pos_db$geometry= NULL
pos_db <- unique(pos_db)

neg_db <- case_int %>%
  select("idunique.1", "match_no")

neg_db$geometry= NULL
neg_db <- unique(neg_db)


head(neg_db)


####Now the samples####
library(tidyverse)
control1 <- neg_db %>% group_by(match_no) %>% sample_n(1)
control2 <- neg_db %>% group_by(match_no) %>% sample_n(2)
control3 <- neg_db %>% group_by(match_no) %>% sample_n(3)
control4 <- neg_db %>% group_by(match_no) %>% sample_n(4)
control5 <- neg_db %>% group_by(match_no) %>% sample_n(5)
control6 <- neg_db %>% group_by(match_no) %>% sample_n(6)
control7 <- neg_db %>% group_by(match_no) %>% sample_n(7)
control8 <- neg_db %>% group_by(match_no) %>% sample_n(8)
control9 <- neg_db %>% group_by(match_no) %>% sample_n(9)
control10<- neg_db %>% group_by(match_no) %>% sample_n(10)