Termplot: A wrapper for 'termplot' that optionally (but by default)...

In Epi: Statistical Analysis in Epidemiology

A wrapper for termplot that optionally (but by default) exponentiates terms, and plot them on a common log-scale. Also scales x-axes to the same physical scale.

Description

The function uses termplot to extract terms from a model with, say, spline, terms, including the standard errors, computes confidence intervals and transform these to the rate / rate-ratio scale. Thus the default use is for models on the log-scale such as Poisson-regression models. The function produces a plot with panels side-by-side, one panel per term, and returns the

Usage

  Termplot( obj,
           plot = TRUE,
           xlab = NULL,
           ylab = NULL,
            xeq = TRUE,
           yshr = 1,
          alpha = 0.05,
          terms = NULL,
         max.pt = NULL )

Arguments

obj	An object with a terms-method — for details the the documentation for termplot.
plot	Should a plot be produced?
xlab	Labels for the x-axes. Defaults to the names of the terms.
ylab	Labels for the x-axes. Defaults to blank.
xeq	Should the units all all plots have the same physical scale for the x-axes).
yshr	Shrinking of y-axis. By default, the y-axes have an extent that accommodates the entire range of confidence intervals. This is a shrinking parameter for the y-axes, setting it to less than 1 will lose a bit of the confidence limits on some of the panels.
alpha	1 minus the confidence level for computing confidence intervals
terms	Which terms should be reported. Passed on to termplot and eventually predict.
max.pt	The maximal number of points in which to report the terms. If NULL all unique points from the analysis dataset are reported for each term (this is a feature of termplot).

Value

A list with one component per term in the model object obj, each component is a 4-column matrix with $x$ as the first column, and 3 columns with estimae and lower and upper confidence limit.

Author(s)

Bendix Cartensen

See Also

Ns, termplot

Examples

# Get the diabetes data and set up as Lexis object
data(DMlate)
DMlate <- DMlate[sample(1:nrow(DMlate),500),]
dml <- Lexis( entry = list(Per=dodm, Age=dodm-dobth, DMdur=0 ),
               exit = list(Per=dox),
        exit.status = factor(!is.na(dodth),labels=c("DM","Dead")),
               data = DMlate )

# Split in 1-year age intervals
dms <- splitLexis( dml, time.scale="Age", breaks=0:100 )

# Model with 6 knots for both age and period
n.kn <- 6
# Model age-specific rates with period referenced to 2004
( a.kn <- with( subset(dms,lex.Xst=="Dead"),
                quantile( Age+lex.dur, probs=(1:n.kn-0.5)/n.kn ) ) )
( p.kn <- with( subset(dms,lex.Xst=="Dead"),
                quantile( Per+lex.dur, probs=(1:n.kn-0.5)/n.kn ) ) )
m2 <- glm( lex.Xst=="Dead" ~ -1 +
                             Ns( Age, kn=a.kn, intercept=TRUE ) +
                             Ns( Per, kn=p.kn, ref=2004 ),
           offset = log( lex.dur ), family=poisson, data=dms )

# Finally we can plot the two effects:
Termplot( m2, yshr=0.9 )

Epi documentation built on Oct. 1, 2024, 5:07 p.m.