[R] xYplot, error bars, log scale

[Jacob Etches]

I'm new to Hmisc and trying to get the following to work, but if I un- 
comment the y-scale list (in order to get a log-scale for the hazard  
ratio), the error bars become strangely large.  The dataframe is  
simply ODS output from TPHREG in SAS.  Can someone point me towards  
what I'm sure is a naive error?  Many thanks!

tphreg1_parameterestimates1987$LCL <-  
exp(tphreg1_parameterestimates1987$estimate - qnorm(. 
975)*tphreg1_parameterestimates1987$stderr)
tphreg1_parameterestimates1987$UCL <-  
exp(tphreg1_parameterestimates1987$estimate + qnorm(. 
975)*tphreg1_parameterestimates1987$stderr)

xYplot(Cbind(exp(estimate),LCL,UCL) ~ lag|malef*ftf, groups=classval0,  
as.table=T,
	ylim=c(.8,4), pch=1:4, type="b", method="alt bars", label.curves=F,  
lty.bar=3,
	data=tphreg1_parameterestimates1987, subset=ft != 2 & parameter ==  
"finc_mean_qu",
	scales=list(
		#y=list(log="e",at=c(1,1.25,1.5,2,2.5,3)),
		x=list(at=seq(3,18,by=3))
		),
	key=list(columns=4,
	    	text=list(c("Q0: Poorest","Q1","Q2","Q3")),
	    	points=list(pch = 1:4,col = trellis.par.get("superpose.symbol") 
$col[1:4])
     		)
	)







Doctoral candidate, Epidemiology Program
Department of Public Health Sciences, University of Toronto Faculty of  
Medicine

Research Associate
Institute for Work & Health
800-481 University Avenue, Toronto, Ontario, Canada   M5G 2E9
T: 416.927.2027 ext. 2290
F: 416.927.4167
jetches at iwh.on.ca
www.iwh.on.ca