求助 R 的大神，求助一个关于改进 R 速度的问题

Or2

2016-04-15 23:15:37 +08:00

data = read.csv("C:/Users/Bamb0o0o0/Desktop/dataset.csv",header = TRUE) # import the data as Data Frame

# dummies

data$logearnings = log(data$earnings)

data$occ1 = ifelse(data$occ == 1,1,0)
data$occ2 = ifelse(data$occ == 2,1,0)
data$occ3 = ifelse(data$occ == 3,1,0)
data$occ4 = ifelse(data$occ == 4,1,0)

data$northeast = ifelse(data$region == 1,1,0)
data$south = ifelse(data$region == 2,1,0)
data$midwest = ifelse(data$region == 3,1,0)
data$west = ifelse(data$region == 4,1,0)
data$central = ifelse(data$region == 5,1,0)

data$exp = data$age - data$edu - 6 # 6 the age people gets in to primary school

data$exp_2 = (data$exp)**2

data$exp_3 = (data$exp)**3

data$edu_ability = data$edu*data$ability

Data=subset(data,select=-c(earnings,occ,region)) # create a new Data Frame with out column earnings, occ, region

female_earnings = subset(Data,Data$female == 1)
male_earnings = subset(Data,Data$female == 0)

############################################################################################
################ Bootstrap ################
############################################################################################

### empty Matrix for coefficients of OLS
matrix_Female = matrix(, nrow = 0, ncol = 11)
matrix_Male = matrix(, nrow = 0, ncol = 11)

for (s in 1:500){
### empty data frame for new sample
df = data.frame(id = integer(), ability = integer(), age = integer(),
female = integer(), edu = integer(), logearnings = double(), occ1 = integer(),
occ2 = integer(), occ3 = integer(), occ4 = integer(), northeast = integer(),
south = integer(), midwest = integer(), west = integer(), central = integer(),
exp = integer(), exp_2 = integer(), exp_3 = integer(), edu_ability = integer())
### create new data frame/ sample)

A = sample(Data$id, 10000, replace = TRUE)

for (i in A) {
df = rbind(df,Data[i,])}

df_female_earnings= subset(df,df$female == 1)
df_male_earnings = subset(df,df$female == 0)

### OLS for new Sample
Model_11_F = lm(df_female_earnings$logearnings ~ df_female_earnings$edu + df_female_earnings$northeast
+ df_female_earnings$south + df_female_earnings$midwest + df_female_earnings$age
+ df_female_earnings$ability + df_female_earnings$exp_2 + df_female_earnings$exp_3
+ df_female_earnings$edu_ability + df_female_earnings$west)
#summary(Model_11_F)
coef(Model_11_F)
matrix_Female=rbind(matrix_Female,coef(Model_11_F))

Model_11_M = lm(df_male_earnings$logearnings ~ df_male_earnings$edu + df_male_earnings$northeast
+ df_male_earnings$south + df_male_earnings$midwest + df_male_earnings$age
+ df_male_earnings$ability + df_male_earnings$exp_2 + df_male_earnings$exp_3
+ df_male_earnings$edu_ability + df_male_earnings$west)
#summary(Model_11_M)
coef(Model_11_M)
matrix_Male=rbind(matrix_Male,coef(Model_11_M))
}

Female_Mean = colMeans(matrix_Female)
Male_Mean = colMeans(matrix_Male)

Matrix_Female_Mean = matrix(colMeans(matrix_Female), nrow=500, ncol=11, byrow=TRUE)
Matrix_Male_Mean = matrix(colMeans(matrix_Male), nrow=500, ncol=11, byrow=TRUE)

#deviation matrix
F = (matrix_Female - Matrix_Female_Mean)^2
M = (matrix_Male - Matrix_Male_Mean)^2

F_sum = colSums(F)
M_sum = colSums(M)

哦，现在贴，第一次发这样的帖子不知道该怎么办。

老师的要就是不让用 boot package ，所以我就索性自己弄了。结果速度慢的一塌糊涂

Or2

2016-04-16 13:36:45 +08:00

data_DF = read.csv("C:/Users/Bamb0o0o0/Desktop/dataset.csv",header = TRUE) # import the data_DF as data_DF Frame

data_DF$logearnings = log(data_DF$earnings)

data_DF$occ1 = ifelse(data_DF$occ == 1,1,0)
data_DF$occ2 = ifelse(data_DF$occ == 2,1,0)
data_DF$occ3 = ifelse(data_DF$occ == 3,1,0)
data_DF$occ4 = ifelse(data_DF$occ == 4,1,0)

data_DF$northeast = ifelse(data_DF$region == 1,1,0)
data_DF$south = ifelse(data_DF$region == 2,1,0)
data_DF$midwest = ifelse(data_DF$region == 3,1,0)
data_DF$west = ifelse(data_DF$region == 4,1,0)
data_DF$central = ifelse(data_DF$region == 5,1,0)

data_DF$exp = data_DF$age - data_DF$edu - 6 # 6 the age people gets in to primary school

data_DF$exp_2 = (data_DF$exp)**2

data_DF$exp_3 = (data_DF$exp)**3

data_DF$edu_ability = data_DF$edu*data_DF$ability

# create a new data_DF Frame with out column earnings, occ, region
data_DF = subset(data_DF,select = -c(earnings,occ,region))

data = data.matrix(data_DF)

BootStrapping = function(){
id = sample(data[,1], 10000, replace = T)
# Data Frame: data_DF$id or data_DF[,1] class(id) now is integer, need less memory
# Matrix: data[,which(col.names(data)== 'id')], class(id) now is numeric, need more memory

New_Sample = data[id,]

# Screening data again
female_earnings = New_Sample[New_Sample[,4] == 1,]
male_earnings = subset(New_Sample, New_Sample[,4] == 0)

# OLS
Model_11_F = lm(female_earnings[,6] ~ female_earnings[,5] + female_earnings[,11]
+ female_earnings[,12] + female_earnings[,13] + female_earnings[,3]
+ female_earnings[,2] + female_earnings[,17] + female_earnings[,18]
+ female_earnings[,19] + female_earnings[,14])

coef(Model_11_F)

Model_11_M = lm(male_earnings[,6] ~ male_earnings[,5] + male_earnings[,11]
+ male_earnings[,12] + male_earnings[,13] + male_earnings[,3]
+ male_earnings[,2] + male_earnings[,17] + male_earnings[,18]
+ male_earnings[,19] + male_earnings[,14])
coef(Model_11_M)

return(c(coef(Model_11_F),coef(Model_11_M)))
}

Numbers_of_Replication = 1000

Coefficient_Matrix = replicate(Numbers_of_Replication,BootStrapping())

# Set Row names
rownames(Coefficient_Matrix) = c('Female_Intercept', 'Female_edu', 'Female_northeast',
'Female_south', 'Female_midwest', 'Female_age', 'Female_ability',
'Female_exp_2', 'Female_exp_3', 'Female_edu_ability', 'Female_west',
'Male_Intercept', 'Male_edu', 'Male_northeast','Male_south', 'Male_midwest',
'Male_age', 'Male_ability', 'Male_exp_2', 'Male_exp_3', 'Male_edu_ability', 'Male_west')

# Create Mean Matrix Residuals and STDER
Coefficient_Mean_Matrix = matrix(rowMeans(Coefficient_Matrix),
nrow = 22, ncol = Numbers_of_Replication, byrow = FALSE)

Residuals_Square = (Coefficient_Matrix - Coefficient_Mean_Matrix)^2

Residuals_Square_Sum = rowSums(Residuals_Square)

STDER = sqrt(1/(Numbers_of_Replication-1)*Residuals_Square_Sum)

这是新改的代码，谁帮我看看还有有什么地方可以改进下