I am trying to confirm that the sample is not different from the population
with respect to geographical respresentation for the following data:
region population sample n_total s_total
1 north 1244 232 9077 1032
2 south 4333 100 9077 1032
3 east 1500 200 9077 1032
4 west 2000 500 9077 1032
For example the first line of this data shows that our sample consisted of
232 individuals from the north that had a population of 1244. The table also
shows that in total, we chose 1032 individuals (s_total) from a population
size of 9077 (n_total).
I thought that one way of doing that would be to run a logistic regression
of sample / population proportion on region (north, south, east, west) and
source (sample, population) and show that the source factor is
insignificant.
I am not exactly satisfied with the results with a huge residual deviance
and a sample factor p value of 1.00.
Therefore, I was wondering if there was a better approach of making such
comparison.
I would appreciate any suggestions.
Raghu
The data / logistic regression code is as below :
____________________________________________________
library(sqldf)
mystring <- ("region, population, sample
north, 1244, 232
south, 4333, 100
east, 1500, 200
west, 2000, 500")
one <- read.table(textConnection(mystring),header=TRUE,sep=",")
n_total <- sum(one$population)
s_total <- sum(one$sample)
one <- cbind(one,n_total,s_total) ; one
two <- sqldf(" select population as count, n_total as total , 'population'
as source, region
from one union select sample as count, s_total as total,
'sample' as source, region
from one " )
attach(two)
prop <- count/total
model1 <- glm(prop ~ source + region, weights=total, quasibinomial)
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
