如何使用 R 调查包分析加权样本中的多个响应问题？

Question

我对 R 比较陌生。我想知道如何使用“调查”包 (http://r-survey.r-forge.r-project.org/survey/) 来分析加权样本的多重响应问题？棘手的一点是，可以勾选多个响应，因此响应存储在多列中。

示例：

我有来自 10 个地区的 500 名受访者的调查数据。假设被问到的主要问题是（存储在 H1_AreYouHappy 列中）：“你快乐吗？” - 是/否/不知道

回答者被问到一个后续问题：“你为什么（不）快乐？” 这是一道选择题，可以勾选多个答案框，因此答案存储在不同的列中，例如：

H1Yes_Why1 (0/1, 即勾选或未勾选) - 'Because of the economny';

H1Yes_Why2 (0/1) - '因为我很健康'；

H1Yes_Why3 (0/1) - '因为我的社交生活'。

这是我的假数据集

districts <- c('Green', 'Red','Orange','Blue','Purple','Grey','Black','Yellow','White','Lavender')
myDataFrame <- data.frame(H1_AreYouHappy=sample(c('Yes','No','Dont Know'),500,rep=TRUE), 
                          H1Yes_Why1 = sample(0:1,500,rep=TRUE), 
                          H1Yes_Why2 = sample(0:1,500,rep=TRUE), 
                          H1Yes_Why3 = sample(0:1,500,rep=TRUE), 
                          District = sample(districts,500,rep=TRUE), stringsAsFactors=TRUE)

我正在使用 R 'survey' 包根据每个地区的实际人口规模应用分层后权重

library(survey)
# Create an unweighted survey object
mySurvey.unweighted <- svydesign(ids=~1, data=myDataFrame)

# Choose which variable contains the sample distribution to be weighted by
sample.distribution <- list(~District)

# Specify (from Census data) how often each level occurs in the population
population.distribution <- data.frame(District = c('Green', 'Red','Orange','Blue','Purple','Grey','Black','Yellow','White','Lavender'),
                              freq = c(0.1824885, 0.0891206, 0.1381343, 0.1006533, 0.1541269, 0.0955853, 0.0268172, 0.0398353, 0.0809459, 0.0922927))

# Apply the weights
mySurvey.rake <- rake(design = mySurvey.unweighted, sample.margins=sample.distribution, population.margins=list(population.distribution))

# Calculate the weighted mean for the main question
svymean(~H1_AreYouHappy, mySurvey.rake)

# How can I calculate the WEIGHTED means for the multiple choice - multiple response follow-up question?

如何计算多项选择题的加权平均值（即跨 0/1 响应列）？

如果我希望它不加权，我可以使用这个函数来计算与我的前缀“H1Yes_Why”匹配的所有列的频率

multipleResponseFrequencies = function(data, question.prefix) {
  # Find the columns with the questions
  a = grep(question.prefix, names(data))
  # Find the total number of responses
  b = sum(data[, a] != 0)
  # Find the totals for each question
  d = colSums(data[, a] != 0)
  # Find the number of respondents
  e = sum(rowSums(data[,a]) !=0)
  # d + b as a vector. This is the overfall frequency 
  f = as.numeric(c(d, b))
  result <- data.frame(question = c(names(d), "Total"),
                       freq = f,
                       percent = (f/b)*100,
                       percentofcases = (f/e)*100)
  result
}
multipleResponseFrequencies(myDataFrame, 'H1Yes_Why')

任何帮助将不胜感激。

Answer 1

我想你想要

svyratio( ~ H1Yes_Why1 + H1Yes_Why2 + H1Yes_Why3 , ~ as.numeric( H1Yes_Why1 + H1Yes_Why2 + H1Yes_Why3 ) , mySurvey.rake)