Below, I first group my data.frame (d) by two categorical variables. First, by gender (2-levels; M/F). Second, by sector (Education, Industry, NGO, Private, Public). Then, I want to sample from each level of sector with the following probabilities: c(.2, .3, .3, .1, .1), and gender by following probabilities c(.4, .6).
I'm using the code below to achieve my goal without success? Is there a fix for that?
Would you please comment if generally my code does what I describe correctly?
d <- read.csv('https://raw.githubusercontent.com/rnorouzian/d/master/su.csv')
library(tidyverse)
set.seed(1)
(out <- d %>%
group_by(gender,sector) %>%
slice_sample(n = 2, weight_by = c(.4, .6, .2, .3, .3, .1, .1))) # `Error: incorrect number of probabilities`
Well slice_sample won't do exactly what you want so I recommend you use splitstackshape to do the job. Install and load as necessary
# install.packages("splitstackshape")
library(splitstackshape)
There are shorter faster ways to specify the proportions table but I'll do it methodically starting from the total sample wanted, which in this case we'll make n = 100 then we'll specify the percentages for the various factor levels.
total_sample <- 100
M_percent <- .4
F_percent <- .6
Education_percent <- .2
Industry_percent <- .3
NGO_percent <- .3
Private_percent <- .1
Public_percent <- .1
Then we call the function stratified with first a vector of the two columns we're operating on, then a vector of groups and the number wanted which we'll calculate from the percentages above...
abc <-
stratified(indt = d,
c("gender", "sector"),
c("F Education" = F_percent * Education_percent * total_sample,
"M Education" = M_percent * Education_percent * total_sample,
"F Industry" = F_percent * Industry_percent * total_sample,
"M Industry" = M_percent * Industry_percent * total_sample,
"F NGO" = F_percent * NGO_percent * total_sample,
"M NGO" = M_percent * NGO_percent * total_sample,
"F Private" = F_percent * Private_percent * total_sample,
"M Private" = M_percent * Private_percent * total_sample,
"F Public" = F_percent * Public_percent * total_sample,
"M Public" = M_percent * Public_percent * total_sample)
)
We get back the randomly selected quantities we requested
head(abc, 20)
fake.name sector pretest state gender pre email phone
1: Correa Education 1254 TX F Medium Correa@...com xxx-xx-1886
2: Manzanares Education 1227 CA F Low Manzanares@...com xxx-xx-1539
3: el-Daoud Education 1409 CA F High el-Daoud@...com xxx-xx-9972
4: Engman Education 1436 CA F High Engman@...com xxx-xx-9446
5: el-Kaba Education 1305 NY F Medium el-Kaba@...com xxx-xx-7060
6: Herrera Education 1405 NY F High Herrera@...com xxx-xx-9146
7: el-Sham Education 1286 TX F Medium el-Sham@...com xxx-xx-4046
8: Harrison Education 1112 NY F Low Harrison@...com xxx-xx-3118
9: Zhu Education 1055 CA F Low Zhu@...com xxx-xx-6223
10: Deguzman Gransee Education 1312 TX F Medium Deguzman Gransee@...com xxx-xx-5676
11: Kearney Education 1303 NY F Medium Kearney@...com xxx-xx-5145
12: Hernandez Mendoza Education 1139 CA F Low Hernandez Mendoza@...com xxx-xx-9642
13: Barros Education 1416 NY M High Barros@...com xxx-xx-2455
14: Torres Education 1370 CA M High Torres@...com xxx-xx-2129
15: King Education 1346 CA M Medium King@...com xxx-xx-5351
16: Cabrera Education 1188 NY M Low Cabrera@...com xxx-xx-6349
17: Lee Education 1208 CA M Low Lee@...com xxx-xx-7713
18: Vernon Education 1216 TX M Low Vernon@...com xxx-xx-7649
19: Ripoll-Bunn Education 1419 TX M High Ripoll-Bunn@...com xxx-xx-8126
20: Ashby Education 1295 TX M Medium Ashby@...com xxx-xx-8416