startEmax | R Documentation |
Compute starting parameter values for iterative procedures for estimating parameters of the 3- or 4- parameter Emax model
startEmax(y,
dose,
baseline,
count=rep(1,length(y)),
modType=3,
binary=FALSE,
lbED50=doselev[2]/10,
ubED50=max(doselev),
lbLambda=0.5,
ubLambda=5)
y |
Outcome (response) variable for the Emax modeling. |
binary |
The default is continuous ( |
dose |
Dose variable corresponding to each outcome value. |
baseline |
Optional baseline covariate(s) of same length as y. When baseline is specified, starting values are created from anacova adjusted dose group means. |
count |
Counts for the number of patients with each dose/y value. Default is 1 (ungrouped data). |
modType |
modType=3 (default) for the 3-parameter Emax model. modType=4 for the 4-parameter Emax model. |
lbED50 |
If the starting ED50 is below lbED50, it is set to lbED50. |
ubED50 |
If the starting ED50 is above ubED50, it is set to ubED50. |
lbLambda |
If the starting lambda is below lbLambda, it is set to lbLambda. |
ubLambda |
If the starting lambda is above ubLambda, it is set to ubLambda. |
Returns a vector with named elements for the starting values for a 3 or 4 parameter Emax model. The order is log(ED50), (lambda, 4 parm), emax, and e0. If baseline is specified, a 'beta' starting parameter is also returned at the end of the vector.
The method is modified from functions created by J. Rogers and start functions supplied with R (SSfp1). The ED50 (and lambda) are computed using the logit-linear relationship between the proportion of the mean response out of the max response and the log(dose). The method assumes placebo data are present, but it will return a starting value even if it is not present. A miniumum of four dose levels is required for 4-parameter starting values.
Neal Thomas
nls
, emaxalt
data("metaData")
exdat<-metaData[metaData$taid==6 & metaData$poptype==1,]
startEmax(exdat$rslt,exdat$dose)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.