计算包含样本设计的分位数（调查包）

Question

我想使用包含复杂调查样本设计的另一列（连续变量）的分位数来计算新列。这个想法是在数据框中创建一个新变量，该变量指示每个观察值属于哪个分位数组

这是我在不包含示例设计的情况下执行该想法的方式，因此您可以了解我的目标。

# Load Data
  data(api)


# Convert data to data.table format (mostly to increase speed of the process)
  apiclus1 <- as.data.table(apiclus1)

# Create deciles variable
apiclus1[, decile:=cut(api00,
                       breaks=quantile(api00,
                                       probs=seq(0, 1, by=0.1), na.rm=T),
                       include.lowest= TRUE, labels=1:10)]

我尝试使用 survey 包中的 svyquantile，但我无法解决这个问题。此代码不会将分位数组作为我可以输入新变量的输出返回。对此有什么想法吗？

# Load Package
 library(survey)

# create survey design
 dclus1 <- svydesign(id=~dnum, weights=~pw, data=apiclus1, fpc=~fpc)

# What I've tried to do
  svyquantile(~api00, design = dclus1, quantiles = seq(0, 1, by=0.1), method = "linear", ties="rounded")

Answer 1

library(survey)

data(api)

dclus1 <- svydesign(id=~dnum, weights=~pw, data=apiclus1, fpc=~fpc)

a <- svyquantile(~api00, design = dclus1, quantiles = seq(0, 1, by=0.1), method = "linear", ties="rounded")

# use factor() and findInterval()
dclus1 <- update( dclus1 , qtile = factor( findInterval( api00 , a ) ) )

# distribution
svymean( ~ qtile , dclus1 )


# or without the one observation in group number 11
dclus1 <- update( dclus1 , qtile = factor( findInterval( api00 , a[ -length( a ) ] ) ) )

# distribution
svymean( ~ qtile , dclus1 )



# quantiles by group
b <- svyby(~api00, ~stype, design = dclus1, svyquantile, quantiles = seq(0, 0.9 , by=0.1) ,ci=T)

# copy over your data
x <- apiclus1

# stype of each record
match( x$stype , b$stype ) 

# create the new qtile variable
x$qtile_by_stype <- factor( diag( apply( data.frame( b )[ match( x$stype , b$stype ) , 2:11 ] , 1 , function( v , w ) findInterval( w , v ) , x$api00 ) ) )

# re-create the survey design
dclus1 <- svydesign(id=~dnum, weights=~pw, data=x, fpc=~fpc)

# confirm you have quantiles
svyby( ~ qtile_by_stype , ~ stype , dclus1 , svymean )

Answer 2

上面整个代码的输出是：

        0   0.1   0.2   0.3   0.4    0.5   0.6    0.7   0.8   0.9   1
api00 411 497.8 535.6 573.2 614.6 651.75 686.6 709.55 735.4 780.7 905

您可以更改名称以代表您的组。 0 和 1 代表最小值和最大值。 0.1 代表十分位 1，0.2 代表十分位 2，等等。类似于：

dt_quantile = svyquantile(~api00, design = dclus1, quantiles = seq(0, 1, by=0.1), method = "linear", ties="rounded")
dt_quantile = data.table(dt_quantile)

setnames(dt_quantile, c("min",paste0("decile",1:10)))

dt_quantile = data.table(t(dt_quantile), keep.rownames = T)

dt_quantile 

#         rn     V1
# 1:      min 411.00
# 2:  decile1 497.80
# 3:  decile2 535.60
# 4:  decile3 573.20
# 5:  decile4 614.60
# 6:  decile5 651.75
# 7:  decile6 686.60
# 8:  decile7 709.55
# 9:  decile8 735.40
# 10: decile9 780.70
# 11: decile10 905.00

我错过了你的目标吗？