Thanks Rui - your initial, very elegant suggestion, has spurred me on!
1. As you noticed, my example data had no examples of duplicate first dates (DOH!)
I have corrected this, and added a test - an ID that has a duplicate which is not the earliest DATE, but is the same DATE an earliest/duplicate for another ID.
2. Your suggestion gave me all the duplicates:
how.many <- ave ( id.d [ ,1], id.d [,1], id.d [,2], FUN = length)
nd.b<- id.d [ how.many > 1, ]
3. I can then simply make a table of earliest DATEs by ID, and then see which DATEs in this table are shared:
earliest <- tapply ( DATE, ID, min)
rownames(earliest[earliest%in%nd.b])
This seems to work - and it does seem exclude IDs which have a duplicate date which is the same as a minimum date for another ID.
I'm trying to work out why!
Many, many thanks for the gift of that function. I will compare the two approaches (and assume that mine is flawed!).
Stuart
************************************************
ID <- c(58,58,58,58,167,167,323,323,323,323,323,323,323
,547,794,814,814,814,814,814,814,841,841,841,841,841
,841,841,841,841,910,910,910,910,910,910,999,1019,1019
,1019)
DATE <-
c(20060821,20061207,20080102,20090904,20040205,20040205,20051111
,20060111,20071119,20080107,20080407,20080521,20080711,20041005
,20070905,20020814,20021125,20040429,20040429,20071205,20080227
,20050421,20050421,20060428,20060602,20060816,20061025,20061129
,20070112,20070514, 19870409,19870508,19870508, 20091120,20091210
,20091224,20050503,19870508,19870508,19880330)
id.d <- cbind (ID,DATE )
how.many <- ave(id.d[,1], id.d[,1], id.d[,2], FUN = length)
nd.b<- id.d[how.many > 1, ]
earliest <- tapply ( DATE, ID, min) # table of earliest DATEs
rownames (earliest [earliest %in% nd.b ] ) # IDs of duplicates at the earliest date for that individual. I think...
******************************************************************
-----Original Message-----
From: Rui Barradas [mailto:ruipbarradas at sapo.pt]
Sent: 23 October 2012 12:21
To: Stuart Leask
Cc: r-help at r-project.org
Subject: Re: [R] [r] How to pick colums from a ragged array?
Hello,
Thinking again, if you just want the first/last in each ID that repeats the DATE, the following function does the job. Since there were no such cases in your data example, I've added 3 rows to the dataset.
ID <- c(58,58,58,58,167,167,323,323,323,323,323,323,323
,547,794,814,814,814,814,814,814,841,841,841,841,841
,841,841,841,841,910,910,910,910,910,910,910,910,999,1019,1019
,1019,1019)
DATE <- c(20060821,20061207,20080102,20090904,20040205,20040323,20051111
,20060111,20071119,20080107,20080407,20080521,20080711,20041005
,20070905,20020814,20021125,20040429,20040429,20071205,20080227
,20050421,20060130,20060428,20060602,20060816,20061025,20061129
,20070112,20070514,20091105,20091105,20091117,20091119,20091120,20091210
,20091224,20091224,20050503,19870508,19880223,19880330,19880330)
id.d <- cbind(ID, DATE)
getRepeat <- function(x, first = TRUE){
fun <- if(first) head else tail
sp <- split(data.frame(x), x[,1])
first.date <- tapply(x[,2], x[,1], FUN = fun, 1)
lst <- lapply(seq_along(sp), function(j) sp[[j]][,2] == first.date[j])
n <- unlist(lapply(lst, sum))
sp1 <- sp[n > 1]
i1 <- lst[n > 1]
lapply(seq_along(sp1), function(j) sp1[[j]][i1[[j]], ]) }
getRepeat(id.d) # defaults to first = TRUE getRepeat(id.d, first = FALSE) # to get the last ones
Hope this helps,
Rui Barradas
Em 23-10-2012 10:59, Rui Barradas escreveu:
Hello,
I'm not sure I understand it well, in the solution below the only
returned value is ID == 814 but it's not the first nor the last DATE.
how.many <- ave(id.d[,1], id.d[,1], id.d[,2], FUN = length)
id.d[how.many > 1, ]
See the help page for ?ave if the repetition of id.d[,1] is confusing.
The first is the vector to average (to apply FUN to) and the second is
one of thw two vectors defining the groups.
Hope this helps,
Rui Barradas
Em 23-10-2012 10:37, Stuart Leask escreveu:
I have a large dataset (~1 million rows) of three variables: ID
(patient's name), DATE (of appointment) and DIAGNOSIS (given on that
date).
Patients may have been assigned more than one diagnosis at any one
appointment - leading to two rows, same ID and DATE but different
DIAGNOSIS.
The diagnoses may change between appointments.
I want to subset the data in two ways:
- define groups of patients by the first diagnosis given
- define groups of patients by the last diagnosis given.
The problem:
Unfortunately, a small number of patients have been given more than
one diagnosis at their first (or last) appointment. These individuals
I need to identify and remove, as it's not possible to say uniquely
what their first (or last) diagnosis was. So I need to identify and
remove these individuals which have pairs of rows with the same ID
and (lowest or highest) DATE. The size of the dataset precludes the
option of doing this by eye.
I suspect there is a very elegant way of doing this in R.
This is what I've come up with:
- Sort by DATE then ID
- Make a ragged array of DATE by ID
- Remove IDs that only occur once.
- Subtract the first and second DATEs. Remove IDs for which
this = zero, as this will only be true for IDs for which the
appointment is recorded twice (because there were two diagnoses
recorded on this date).
- (Then do the same to get the 'last appointment'
duplicates, by reversing the initial sort by DATE.)
I am stuck at the 'Subtract dates' step: I would like to get the data
out of the ragged array by columns (so e.g. I end up with a matrix of
ID, 1st DATE, 2nd DATE). But I can't get the dates out by column from
the ragged array.
I hope someone can help. My ugly code is below, with some data for
testing.
Stuart
Dr Stuart John Leask DM FRCPsych MB BChir MA Clinical Senior Lecturer
and Honorary Consultant Pychiatrist Institute of Mental Health,
Innovation Park Triumph Road, Nottingham, Notts. NG7 2TU. UK Tel. +44
115 82 30419
stuart.leask at nottingham.ac.uk<mailto:stuart.leask at nottingham.ac.uk>
Google 'Dr Stuart Leask'
ID <- c(58,58,58,58,167,167,323,323,323,323,323,323,323
,547,794,814,814,814,814,814,814,841,841,841,841,841
,841,841,841,841,910,910,910,910,910,910,999,1019,1019
,1019)
DATE <-
c(20060821,20061207,20080102,20090904,20040205,20040323,20051111
,20060111,20071119,20080107,20080407,20080521,20080711,20041005
,20070905,20020814,20021125,20040429,20040429,20071205,20080227
,20050421,20060130,20060428,20060602,20060816,20061025,20061129
,20070112,20070514,20091105,20091117,20091119,20091120,20091210
,20091224,20050503,19870508,19880223,19880330)
id.d <- cbind (ID,DATE )
rag.a <- split ( id.d [ ,2 ], id.d [ ,1]) # create
ragged array, 1-n DATES for every NAME
# Inelegant attempt to remove IDs that only have one entry:
rag.s <-tapply (id.d [ ,2], id.d [ ,1], sum) #add up the dates per
row # Since DATE is in 'year mo da', if there's only one date, sum
will be less than 2100000:
rag.t <- rag.s [ rag.s > 21000000 ]
multi.dates <- rownames ( rag.t ) # all the
IDs with >1 date
rag.am <- rag.a [ multi.dates ] # rag.am
only has IDs with > 1 Date
# But now I'm stuck.
# Each row of the array is rag.am$ID.
# So I can't pick columns of DATEs from the ragged array.
This message and any attachment are intended solely for the addressee
and may contain confidential information. If you have received this
message in error, please send it back to me, and immediately delete
it. Please do not use, copy or disclose the information contained
in this message or in any attachment. Any views or opinions
expressed by the author of this email do not necessarily reflect the
views of the University of Nottingham.
This message has been checked for viruses but the contents of an
attachment may still contain software viruses which could damage your
computer
system:
you are advised to perform your own checks. Email communications with
the University of Nottingham may be monitored as permitted by UK
legislation.
[[alternative HTML version deleted]]