Hi,
This gives me more combinations than you got with SAS. Also, this selects the
one with minimum dimension between duplicates.
final3New<-read.table(file="real_data_cecilia.txt",sep="\t",header=T)
final3New1<-read.csv("real_data_cecilia_new.csv")
fun3<- function(dat){
if(any(duplicated(dat))){
indx<- which(duplicated(dat))
row.names(dat)<-1:nrow(dat)
dat1<- subset(dat[indx,],dummy==1)
dat0<- subset(dat[indx,],dummy==0)
indx1<- as.numeric(row.names(dat1))
indx11<- sort(c(indx1,indx1+1))
indx0<- as.numeric(row.names(dat0))
indx00<- sort(c(indx0,indx0-1))
indx10<- sort(c(indx11,indx00))
res <- dat[-indx10,]
res
}
else {
dat
}
}
fun1New<-function(dat,percent,number){
lst1<- split(dat,list(dat$year,dat$industry))
lst2<- lst1[lapply(lst1,nrow)>1]
lst3<- lapply(lst2,function(x) {
CombN1<-combn(seq_len(nrow(x)),2)
lapply(split(CombN1,col(CombN1)),function(y){
x1<-x[y,]
x1[sum(x1$dummy)==1,]
})
})
lst4<- lapply(lst3,function(x) x[lapply(x,nrow)>0])
lst5<- lst4[lapply(lst4,length)>0]
lst6<- lapply(lst5,function(x){
lapply(x,function(y){
x1<- abs(diff(y$dimension))< number
x2<- y$dimension[2]+ (y$dimension[2]*percent)
x3<- y$dimension[2]- (y$dimension[2]*percent)
x4<- y$dimension[1]+ (y$dimension[1]*percent)
x5<- y$dimension[1]- (y$dimension[1]*percent)
x6<- (y$dimension[1] < x2) & (y$dimension[1] > x3)
x7<- (y$dimension[2]< x4) & (y$dimension[2]> x5)
y[((x6 & x1)| (x7 & x1)),]
})
}
)
lst7<- lapply(lst6,function(x) x[lapply(x,nrow)>0])
lst8<- lst7[lapply(lst7,length)>0]
lst9<- lapply(lst8,function(x) do.call(rbind,x))
lst10<-lapply(lst9,function(x) {
row.names(x)<- 1:nrow(x)
x1<-x[x$dummy==1,]
do.call(rbind,lapply(split(x1,x1$dimension),function(y){
indx1<-sort(c(as.numeric(row.names(y)),as.numeric(row.names(y))+1))
x2<-x[indx1,]
x3<- subset(x2,dummy==0)
x4<-x3[which.min(abs(x2$dimension[1]-x3$dimension)),]
rbind(x2[1,],x4)
}))
})
res<- do.call(rbind,lapply(lst10,fun3))
row.names(res)<- 1:nrow(res)
res
}
####1st dataset
res10PercentHigh<- fun1New(final3New,0.10,500000000)
dim(res10PercentHigh)
#[1] 764 5
dim(unique(res10PercentHigh))
#[1] 764 5
nrow(subset(res10PercentHigh,dummy==0))
#[1] 382
nrow(subset(res10PercentHigh,dummy==1))
#[1] 382
res10PercentLow<- fun1New(final3New,0.10,50)
dim(res10PercentLow)
#[1] 294 5
dim(unique(res10PercentLow))
#[1] 294 5
nrow(subset(res10PercentLow,dummy==0))
#[1] 147
nrow(subset(res10PercentLow,dummy==1))
#[1] 147
res5PercentHigh<- fun1New(final3New,0.05,500000000)
dim(res5PercentHigh)
#[1] 630 5
dim(unique(res5PercentHigh))
#[1] 630 5
nrow(subset(res5PercentHigh,dummy==0))
#[1] 315
nrow(subset(res5PercentHigh,dummy==1))
#[1] 315
res5PercentLow<- fun1New(final3New,0.05,50)
dim(res5PercentLow)
#[1] 294 5
dim(unique(res5PercentLow))
#[1] 294 5
nrow(subset(res5PercentLow,dummy==0))
#[1] 147
nrow(subset(res5PercentLow,dummy==1))
#[1] 147
#######2nd dataset
res10PercentHigh<- fun1New(final3New1,0.10,500000000)
dim(res10PercentHigh)
#[1] 760 5
dim(unique(res10PercentHigh))
#[1] 760 5
nrow(subset(res10PercentHigh,dummy==0))
#[1] 380
nrow(subset(res10PercentHigh,dummy==1))
#[1] 380
res10PercentLow<- fun1New(final3New1,0.10,100)
dim(res10PercentLow)
#[1] 418 5
dim(unique(res10PercentLow))
#[1] 418 5
nrow(subset(res10PercentLow,dummy==0))
#[1] 209
nrow(subset(res10PercentLow,dummy==1))
#[1] 209
res5PercentHigh<- fun1New(final3New1,0.05,500000000)
dim(res5PercentHigh)
#[1] 640 5
dim(unique(res5PercentHigh))
#[1] 640 5
nrow(subset(res5PercentHigh,dummy==0))
#[1] 320
nrow(subset(res5PercentHigh,dummy==1))
#[1] 320
res5PercentLow<- fun1New(final3New1,0.05,50)
dim(res5PercentLow)
#[1] 310 5
dim(unique(res5PercentLow))
#[1] 310 5
nrow(subset(res5PercentLow,dummy==0))
#[1] 155
nrow(subset(res5PercentLow,dummy==1))
#[1] 155
res20PercentHigh<- fun1New(final3New1,0.20,500000000)
dim(res20PercentHigh)
#[1] 846 5
dim(unique(res20PercentHigh))
#[1] 846 5
nrow(subset(res20PercentHigh,dummy==0))
#[1] 423
nrow(subset(res20PercentHigh,dummy==1))
#[1] 423
A.K.
----- Original Message -----
From: Cecilia Carmo <cecilia.ca...@ua.pt>
To: arun <smartpink...@yahoo.com>
Cc:
Sent: Sunday, June 16, 2013 5:57 AM
Subject: RE: matched samples, dataframe, panel data
In the script I send you and with the file that I sent with it and with the
old function 1 and 2
it got 350 combinations and it was possible to have more
Now with new fun 1 and 3 I have less, so it is not ok, does it?
> res10Percent<- fun1New(final3New2,0.10,500000000)
> res10F3<- fun3(res10Percent)
> dim(res10F3)
[1] 600 5
> nrow(subset(res10F3,dummy==0))
[1] 300
> nrow(subset(res10F3,dummy==1))
[1] 300
Sorry for making you spending so much time. I thought it could be easier.
Cecília
________________________________________
De: arun [smartpink...@yahoo.com]
Enviado: sexta-feira, 14 de Junho de 2013 23:09
Para: Cecilia Carmo
Assunto: Re: matched samples, dataframe, panel data
One thing I forgot to mention. I used fun3() because i found fun2() still have
some problems with getting the correct dimensions. You can check the results
of fun1() and fun3() and see if all the combinations are got. Then, if I get
chance, I will correct fun2().
"""""
And you conclude that they are the same!
"""""""
Here, also I am not concluding anything.
A.K.
----- Original Message -----
From: arun <smartpink...@yahoo.com>
To: Cecilia Carmo <cecilia.ca...@ua.pt>
Cc: R help <r-help@r-project.org>
Sent: Friday, June 14, 2013 6:05 PM
Subject: Re: matched samples, dataframe, panel data
Hi,
I changed the fun1(). Now, it should be possible to get all the possible
combinations within each group.
final3New<-read.table(file="real_data_cecilia.txt",sep="\t",header=T)
final3New1<-read.csv("real_data_cecilia_new.csv")
fun1New<- function(dat,percent,number){
lst1<- split(dat,list(dat$year,dat$industry))
lst2<- lst1[lapply(lst1,nrow)>1]
lst3<- lapply(lst2,function(x) {
CombN1<-combn(seq_len(nrow(x)),2)
lapply(split(CombN1,col(CombN1)),function(y){
x1<-x[y,]
x1[sum(x1$dummy)==1,]
})
})
lst4<- lapply(lst3,function(x) x[lapply(x,nrow)>0])
lst5<- lst4[lapply(lst4,length)>0]
lst6<- lapply(lst5,function(x){
lapply(x,function(y){
x1<- abs(diff(y$dimension))< number
x2<- y$dimension[2]+ (y$dimension[2]*percent)
x3<- y$dimension[2]- (y$dimension[2]*percent)
x4<- (y$dimension[1] < x2) & (y$dimension[1] > x3)
y[x4 & x1,]
})
}
)
lst7<- lapply(lst6,function(x) x[lapply(x,nrow)>0])
lst8<- lst7[lapply(lst7,length)>0]
res<- do.call(rbind,lapply(lst8,function(x){
do.call(rbind,x)
}))
row.names(res)<- 1:nrow(res)
res
}
##Applying fun1New
res5Percent<- fun1New(final3New,0.05,50)
dim(res5Percent)
#[1] 718 5
res5PercentHigh<- fun1New(final3New,0.05,500000)
dim(res5PercentHigh)
#[1] 2788 5
res5Percent1<- fun1New(final3New1,0.05,50)
dim(res5Percent1)
#[1] 870 5
res5Percent1High<- fun1New(final3New1,0.05,500000)
dim(res5Percent1High)
#[1] 2902 5
res10Percent<- fun1New(final3New,0.10,200)
dim(res10Percent)
#[1] 2928 5
res10Percent1<- fun1New(final3New1,0.10,200)
dim(res10Percent1)
#[1] 3092 5
fun3<- function(dat){
indx<- duplicated(dat)
dat1<- subset(dat[indx,],dummy==1)
dat0<- subset(dat[indx,],dummy==0)
indx1<- as.numeric(row.names(dat1))
indx11<- sort(c(indx1,indx1+1))
indx0<- as.numeric(row.names(dat0))
indx00<- sort(c(indx0,indx0-1))
indx10<- sort(c(indx11,indx00))
res <- dat[-indx10,]
res
}
#Applying fun3()
res5F3<- fun3(res5Percent)
dim(res5F3)
#[1] 278 5
res5F3High<- fun3(res5PercentHigh)
dim(res5F3High)
#[1] 546 5
res5F3_1<- fun3(res5Percent1)
#[1] 302 5
res5F3High_1<- fun3(res5Percent1High)
dim(res5F3High_1)
#[1] 570 5
res10F3<- fun3(res10Percent)
dim(res10F3)
#[1] 462 5
res10F3_1<- fun3(res10Percent1)
#[1] 474 5
nrow(subset(res5F3,dummy==0))
#[1] 139
nrow(subset(res5F3,dummy==1))
#[1] 139
nrow(subset(res5F3High,dummy==1))
#[1] 273
nrow(subset(res5F3High,dummy==0))
#[1] 273
nrow(subset(res10F3,dummy==0))
#[1] 231
nrow(subset(res10F3,dummy==1))
#[1] 231
nrow(subset(res10F3_1,dummy==1))
#[1] 237
nrow(subset(res10F3_1,dummy==0))
#[1] 237
dim(unique(res5F3))
#[1] 278 5
dim(unique(res5F3High))
#[1] 546 5
dim(unique(res10F3_1))
#[1] 474 5
dim(unique(res10F3))
#[1] 462 5
A.K.
________________________________
From: Cecilia Carmo <cecilia.ca...@ua.pt>
To: arun <smartpink...@yahoo.com>
Sent: Friday, June 14, 2013 10:44 AM
Subject: me again
There some matchs that are missing. That is, it is possible to have more matchs.
I'm sending you a sript and the data.
Than you.
Cecília
______________________________________________
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.
