A cheatsheet for R.

See on CiteULike.

> install.packages("PKGNAME", dependencies=TRUE)

Seen there.

DF<-DF[order(DF[,1]),]

or

DF<-DF[order(DF$column),]

Reduce the number of rows in a data frame to a given length, with a uniform distribution throughout the initial frame.

filter.subsample <- function(dataframe, maxelt) dataframe[seq(1,(nrow(dataframe)),max(1,nrow(dataframe)/maxelt)),]

Or more easily,

dataframe[sample(1,nrow(dataframe)), maxelt]

for loops seem to be very costly and not optimised. It appears to be more efficient to manipulate full datasets at once.

Let's assume a dataframe as follows.

DF <- data.frame(time=jitter(1:400), gauge=randu[,1])

The idea is to have temporary vectors of time and absolute data with their index offset by one, then subtract and divide appropriately, and fill in with NAs

size = nrow(DF)
time1 = c(DF$time[2:size], NA)
gauge1 = c(DF$gauge[2:size], NA)
DF$rate <- (gauge1 - DF$gauge) / (time1 - DF$time)

Four graphs-page, two on top and a large one at the bottom.

layout(matrix(c(1,2,3,3), 2, 2, byrow = TRUE))

Seen here.

## add `drop.unused.levels' argument to boxplot.formula
boxplot2 <-
function (formula, data = NULL, ...,
           subset, na.action = NULL,
           drop.unused.levels = TRUE)
{
     if (missing(formula) || (length(formula) != 3))
         stop("'formula' missing or incorrect")
     m <- match.call(expand.dots = FALSE)
     if (is.matrix(eval(m$data, parent.frame())))
         m$data <- as.data.frame(data)
     m$... <- NULL
     m$na.action <- na.action
     m$drop.unused.levels <- drop.unused.levels
     m[[1]] <- as.name("model.frame")
     mf <- eval(m, parent.frame())
     response <- attr(attr(mf, "terms"), "response")
     boxplot(split(mf[[response]], mf[-response]), ...)
}

library(RSQLite)
con <- dbConnect(drv="SQLite", dbname="db.sq3")
d <- dbGetQuery(con, "select ...")

Seen here.

ISOdatetime(1970,1,1,0,0,0) + unixtime