planMod {DoseFinding}  R Documentation 
Evaluate performance metrics for fitting doseresponse models
Description
This function evaluates, the performance metrics for fitting doseresponse models (using asymptotic approximations or simulations). Note that some metrics are available via the print method and others only via the summary method applied to planMod objects. The implemented metrics are
Root of the meansquared error to estimate the placeboadjusted doseresponse averaged over the used doselevels, i.e. a rather discrete set (
dRMSE
). Available via the print method of planMod objects.Root of the meansquared error to estimate the placeboadjusted doseresponse (
cRMSE
) averaged over fine (almost continuous) grid at 101 equally spaced values between placebo and the maximum dose. NOTE: Available via the summary method applied to planMod objects.Ratio of the placeboadjusted meansquared error (at the observed doses) of modelbased vs ANOVA approach (
EffvsANOVA
). This can be interpreted on the sample size scale. NOTE: Available via the summary method applied to planMod objects.Power that the (unadjusted) onesided ‘1alpha’ confidence interval comparing the dose with maximum effect vs placebo is larger than ‘tau’. By default ‘alpha = 0.025’ and ‘tau = 0’ (
Pow(maxDose)
). Available via the print method of planMod objects.Probability that the EDp estimate is within the true [EDpLB, EDpUB] (by default ‘p=0.5’, ‘pLB=0.25’ and ‘pUB=0.75’). This metric gives an idea on the ability to characterize the increasing part of the doseresponse curve (
P(EDp)
). Available via the print method of planMod objects.Length of the quantile range for a target dose (TD or EDp). This is calculated by taking the difference of the dUB and dLB quantile of the empirical distribution of the dose estimates. (
lengthTDCI
andlengthEDpCI
). It is NOT calculated by calculating confidence interval lengths in each simulated dataset and taking the mean. NOTE: Available via the summary method of planMod objects.
A plot method exists to summarize doseresponse and dose estimations graphically.
Usage
planMod(model, altModels, n, sigma, S, doses, asyApprox = TRUE,
simulation = FALSE, alpha = 0.025, tau = 0, p = 0.5,
pLB = 0.25, pUB = 0.75, nSim = 100, cores = 1,
showSimProgress = TRUE, bnds, addArgs = NULL)
## S3 method for class 'planMod'
plot(x, type = c("doseresponse", "ED", "TD"),
p, Delta, placAdj = FALSE, xlab, ylab, ...)
## S3 method for class 'planMod'
summary(object, digits = 3, len = 101,
Delta, p, dLB = 0.05, dUB = 0.95, ...)
Arguments
model 
Character vector determining the doseresponse model(s) to be used for fitting the data. When more than one doseresponse model is provided the best fitting model is chosen using the AIC. Builtin models are "linlog", "linear", "quadratic", "emax", "exponential", "sigEmax", "betaMod" and "logistic" (see drmodels). 
altModels 
An object of class ‘Mods’, defining the true mean vectors under which operating characteristics should be calculated. 
n , sigma , S 
Either a vector ‘n’ and ‘sigma’ or ‘S’ need to be
specified. When ‘n’ and ‘sigma’ are specified it is
assumed computations are made for a normal homoscedastic ANOVA model
with group sample sizes given by ‘n’ and residual standard
deviation ‘sigma’, i.e. the covariance matrix used for the
estimates is thus When ‘S’ is specified this will be used as covariance matrix for the estimates. 
doses 
Doses to use 
asyApprox , simulation 
Logicals determining, whether asymptotic approximations or simulations should be calculated. If multiple models are specified in ‘model’ asymptotic approximations are not available. 
alpha , tau 
Significance level for the onesided confidence interval for modelbased contrast of best dose vs placebo. Tau is the threshold to compare the confidence interval limit to. CI(MaxDCont) gives the percentage that the bound of the confidence interval was larger than tau. 
p , pLB , pUB 
p determines the type of EDp to estimate. pLB and pUB define the bounds for the EDp estimate. The performance metric Pr(IdED) gives the percentage that the estimated EDp was within the true EDpLB and EDpUB. 
nSim 
Number of simulations 
cores 
Number of cores to use for simulations. By default 1 cores is used, note that cores > 1 will have no effect Windows, as the mclapply function is used internally. 
showSimProgress 
In case of simulations show the progress using a progressbar. 
bnds 
Bounds for nonlinear parameters. This needs to be a list with list
entries corresponding to the selected bounds. The names of the list
entries need to correspond to the model names. The

addArgs 
See the corresponding argument in function

x 
An object of class planMod 
type 
Type of plot to produce 
Delta 
Additional arguments determining what dose estimate to plot, when ‘type = "ED"’ or ‘type = "TD"’ 
placAdj 
When ‘type = "doseresponse"’, this determines whether doseresponse estimates are shown on placeboadjusted or original scale 
xlab , ylab 
Labels for the plot (ylab only applies for ‘type = "doseresponse"’) 
len 
Number of equally spaced points to determine the meansquared error on a grid (cRMSE). 
dLB , dUB 
Which quantiles to use for calculation of 
object , digits 
object: A planMod object. digits: Digits in summary output 
... 
Additional arguments (currently ignored) 
Author(s)
Bjoern Bornkamp
References
TBD
See Also
Examples
## Not run:
doses < c(0,10,25,50,100,150)
fmodels < Mods(linear = NULL, emax = 25,
logistic = c(50, 10.88111), exponential= 85,
betaMod=rbind(c(0.33,2.31),c(1.39,1.39)),
doses = doses, addArgs=list(scal = 200),
placEff = 0, maxEff = 0.4)
sigma < 1
n < rep(62, 6)*2
model < "quadratic"
pObj < planMod(model, fmodels, n, sigma, doses=doses,
simulation = TRUE,
alpha = 0.025, nSim = 200,
p = 0.5, pLB = 0.25, pUB = 0.75)
print(pObj)
## to get additional metrics (e.g. EffvsANOVA, cRMSE, lengthTDCI, ...)
summary(pObj, p = 0.5, Delta = 0.3)
plot(pObj)
plot(pObj, type = "TD", Delta=0.3)
plot(pObj, type = "ED", p = 0.5)
## End(Not run)