An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120719/baf793b0/attachment.pl>
Subsetting problem data, 2
2 messages · Lib Gray, Rui Barradas
Hello,
Try the following. The data is your example of Patient A through E, but
from the output of dput().
dat <- structure(list(Patient = structure(c(1L, 1L, 1L, 1L, 1L, 2L,
2L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 4L, 5L, 5L, 5L), .Label = c("A",
"B", "C", "D", "E"), class = "factor"), Cycle = c(1L, 2L, 3L,
4L, 5L, 1L, 2L, 1L, 3L, 4L, 5L, 1L, 2L, 4L, 5L, 1L, 2L, 3L),
V1 = c(0.4, 0.3, 0.3, 0.4, 0.5, 0.4, 0.4, 0.9, 0.3, NA, 0.4,
0.2, 0.5, 0.6, 0.5, 0.1, 0.5, 0.4), V2 = c(0.1, 0.2, NA,
NA, 0.2, NA, NA, 0.9, 0.5, NA, NA, 0.5, 0.7, 0.4, 0.5, NA,
0.3, 0.3), V3 = c(0.5, 0.5, 0.6, 0.4, 0.5, NA, NA, 0.9, 0.6,
NA, NA, NA, NA, NA, NA, NA, NA, NA), V4 = c(1.5, 1.6, 1.7,
1.8, 1.5, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA,
NA), V5 = c(NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA,
NA, NA, NA, NA, NA, NA)), .Names = c("Patient", "Cycle",
"V1", "V2", "V3", "V4", "V5"), class = "data.frame", row.names = c(NA,
-18L))
dat
nms <- names(dat)[grep("^V[1-9]$", names(dat))]
dd <- split(dat, dat$Patient)
fun <- function(x) any(is.na(x)) && any(!is.na(x))
ix <- sapply(dd, function(x) Reduce(`|`, lapply(x[, nms], fun)))
dd[ix]
do.call(rbind, dd[ix])
I'm assuming that the variables names are as posted, V followed by one
single digit 1-9. To keep the Patients with complete cases just negate
the index 'ix', it's a logical index.
Note also that dput() is the best way of posting a data example.
Hope this helps,
Rui Barradas
Em 19-07-2012 15:15, Lib Gray escreveu:
Hello, I didn't give enough information when I sent an query before, so I'm trying again with a more detailed explanation: In this data set, each patient has a different number of measured variables (they represent tumors, so some people had 2 tumors, some had 5, etc). The problem I have is that often in later cycles for a patient, tumors that were originally measured are now missing (or a "new" tumor showed up). We assume there are many different reasons for why a tumor would be measured in one cycle and not another, and so I want to subset OUT the "problem" patients to better study these patterns. An example: Patient Cycle V1 V2 V3 V4 V5 A 1 0.4 0.1 0.5 1.5 NA A 2 0.3 0.2 0.5 1.6 NA A 3 0.3 NA 0.6 1.7 NA A 4 0.4 NA 0.4 1.8 NA A 5 0.5 0.2 0.5 1.5 NA I want to keep patient A; they have 4 measured tumors, but tumor 2 is missing data for cycles 3 and 4 B 1 0.4 NA NA NA NA B 2 0.4 NA NA NA NA I do not want to keep patient B; they have 1 tumor that is measure consistently in both cycles C 1 0.9 0.9 0.9 NA NA C 3 0.3 0.5 0.6 NA NA C 4 NA NA NA NA NA C 5 0.4 NA NA NA NA I do want to keep patient C; all their data is missing for cycle 4 and cycle 5 only measured one tumor D 1 0.2 0.5 NA NA NA D 2 0.5 0.7 NA NA NA D 4 0.6 0.4 NA NA NA D 5 0.5 0.5 NA NA NA I do not want patient D, their two tumors were measured each cycle E 1 0.1 NA NA NA NA E 2 0.5 0.3 NA NA NA E 3 0.4 0.3 NA NA NA I DO want patient E; they only had one tumor register in Cycle 1, but cycles 2 and 3 had two tumors. Thanks for any help! [[alternative HTML version deleted]]
______________________________________________ R-help at 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.