Hmisc (version 3.0-1)

Misc: Miscellaneous Functions

Description

This documents miscellaneous small functions in Hmisc that may be of interest to users.

confbar draws multi-level confidence bars using small rectangles that may be of different colors.

james.stein computes James-Stein shrunken estimates of cell means given a response variable (which may be binary) and a grouping indicator.

km.quick provides a fast way to invoke survfit.km in the survival package to get Kaplan-Meier estimates for a single stratum for a vector of time points (if times is given) or to get a vector of survival time quantiles (if q is given).

lm.fit.qr.bare is a fast stripped-down function for computing regression coefficients, residuals, $R^2$, and fitted values. It uses the Fortran routines dqrls.

matxv multiplies a matrix by a vector, handling automatic addition of intercepts if the matrix does not have a column of ones. If the first argument is not a matrix, it will be converted to one.

nomiss returns a data frame (if its argument is one) with rows corresponding to NAs removed, or it returns a matrix with rows with any element missing removed.

outerText uses text() to put test strings in left or right margins. It temporarily sets par(xpd=NA) if using R.

sepUnitsTrans converts character vectors containing values such as c("3 days","3day","4month","2 years","2weeks","7") to numeric vectors (here c(3,3,122,730,14,7)) in a flexible fashion. The user can specify a vector of units of measurements and conversion factors. The units with a conversion factor of 1 are taken as the target units, and if those units are present in the character strings they are ignored. The target units are added to the resulting vector as the "units" attribute. trap.rule computes the area under a curve using the trapezoidal rule, assuming x is sorted.

trellis.strip.blank sets up Trellis or Lattice graphs to have a clear background on the strips for panel labels. under.unix is a scalar logical value that is TRUE if you are running Linux or Unix.

.R. is a logical value set to TRUE if running R, FALSE for S-Plus.

.SV4. is a logical value set to TRUE if running version 4 of the S language under S-Plus, FALSE otherwise. unPaste provides a version of the S-Plus unpaste that works for Rand S-Plus.

whichClosePW is a very fast function using weighted multinomial sampling to determine which element of a vector is "closest" to each element of another vector. whichClosest quickly finds the closest element without any randomness.

xless is a function for Linux/Unix users to invoke the system xless command to pop up a window to display the result of printing an object.

Usage

confbar(at, est, se, width, q = c(0.7, 0.8, 0.9, 0.95, 0.99), 
        col = if (.R.) gray(c(0, 0.25, 0.5, 0.75, 1)) else
              if (under.unix) c(1, 0.8, 0.5, 0.2, 0.065) else
              c(1, 4, 3, 2, 5),
        type = c("v", "h"), labels = TRUE, ticks = FALSE,
        cex = 0.5, side = "l", lwd = 5, clip = c(-1e+30, 1e+30),
        fun = function(x) x,
        qfun = function(x) ifelse(x == 0.5, qnorm(x),
                            ifelse(x < 0.5, qnorm(x/2),
                            qnorm((1 +  x)/2))))
james.stein(y, group)
km.quick(S, times, q)
lm.fit.qr.bare(x, y, tolerance, intercept=TRUE, xpxi=FALSE)
matxv(a, b, kint=1)
nomiss(x)
outerText(string, y, setAside=string[1], side=4, space=1,
          adj=1, cex=par('cex'))
sepUnitsTrans(x, conversion=c(day=1, month=365.25/12, year=365.25, week=7),
              round=FALSE, digits=0)
trap.rule(x, y)
trellis.strip.blank()
unPaste(str, sep="/", extended=FALSE)
whichClosest(x, w)
whichClosePW(x, w, f=0.2)
xless(x, ..., title)

Arguments

at

x-coordinate for vertical confidence intervals, y-coordinate for horizontal

est

vector of point estimates for confidence limits

se

vector of standard errors

width

width of confidence rectanges in user units

q

vector of confidence coefficients or quantiles

col

vector of colors

type

"v" for vertical, "h" for horizontal

labels

set to FALSE to omit drawing confidence coefficients

ticks

set to TRUE to draw lines between rectangles

cex

character expansion factor

side

for confbar is "b","l","t","r" for bottom, left, top, right. For outText is the using integers 1-4 corresponding to these.

lwd

line widths

clip

interval to truncate limits

fun

function to transform scale

qfun

quantiles on transformed scale

group

a categorical grouping variable

S

a Surv object

times

a numeric vector of times

tolerance

tolerance for judging singularity in matrix

intercept

set to FALSE to not automatically add a column of ones to the x matrix

xpxi

set to TRUE to add an element to the result containing the inverse of $X'X$

a

a numeric matrix or vector

b

a numeric vector

kint

which element of b to add to the result if a does not contain a column for intercepts

string

a character string vector

setAside

for adj=1 side=4, is a character string used to determine the space to set aside for all strings.

space

the number of extra characters to leave to the left of the string(s) (adj=0) or to the right (adj=1)

adj

0 for left justification, 0.5 for centered, 1 for right

conversion

a named numeric vector

round

set to TRUE to round converted values

digits

number of digits used for round

str

a character string vector

w

a numeric vector

x

a numeric vector (matrix for lm.fit.qr.bare) or data frame. For xless may be any object that is sensible to print. For sepUnitsTrans is a character or factor variable.

y

a numeric vector

sep

a single character string specifying the delimiter

extended

see strsplit in R

f

a scaling constant

title

a character string to title a window or plot

...

arguments passed through to another function

Examples

trap.rule(1:100,1:100)

unPaste(c('a;b or c','ab;d','qr;s'), ';')

sepUnitsTrans(c('3 days','4 months','2 years','7'))

set.seed(1)
whichClosest(1:100, 3:5)
whichClosest(1:100, rep(3,20))

whichClosePW(1:100, rep(3,20))
whichClosePW(1:100, rep(3,20), f=.05)
whichClosePW(1:100, rep(3,20), f=1e-10)