############################################### # HOW TO USE THIS FILE # This is an HTML version of the R script file ex4.R # It is intended to show you the code and to allow links. # NOT to use interactively to run the code # For that download the script file ex4.R # to view the ouptut from these commands in html go to R results page
red stuff (R commands)
# # Links in this page Equal opportunities policies weighted and unweighted - Table 4.3 E O policy by workplace size Table 4.4 Other factors by equal opportunities policy Table 4.5 Logistic regression Table 4.6 Logistic regression Table 4.7 Finite population corrections Table 4.8 # to get names of variables names(wers) # # the ethnic group percentages are in a column labelled ETHNIC # to get a histogram of this variable do # hist(wers$ETHNIC) # # now it is time to try using the survey package # to give R access to the survey functions do - see mini guide if this fails # library(survey) Back to top## ################ table 4.3 ############################### # # # The first step is to create a SURVEY DESIGN OBJECT that # will contain both the data and the information about the design # # This design has WEIGHTS (EST_WT),but to get the correct # design effects without using the option deff="replace" # we need to use the weights that add to the population (GROSSWT) # and it is stratified by size of workplace # and sector. There are 71 strata indexed by STRATA # Since it is not clustered we set the sampling unit ids # to be the serial number SERNO, different for each workplace wers.des <- svydesign(id=~SERNO, strata=~STRATA,weights=~GROSSWT,data=wers) # # # now we can look at the survey weighted mean of the proportions # of workplaces with an EO policy using the svymean command # giving the design as the second argument to the function # svymean(~EO,wers.des) # # sadly this just gives a missing value since EO has 7 missing values # but the answer is to remove the missing data from the survey design # see the help for is.na and for subset.surveydesign # svymean(~EO,subset(wers.des,!is.na(EO)),deff=T) # # we can compare this with the unweighted mean had the design been different # wers.des <- svydesign(id=~SERNO, strata=~STRATA,data=wers) svymean(~EO,subset(wers.des,!is.na(EO)),deff=T) # Back to top