Dear All,
Can we generate data sets from n original data sets (bootstrap) using Phoenix NLME ??
Best regards,
Raghav
Dear Raghav
If your question is if we can simulate datasets from n original data sets that are appended into one global data set, then the answer is not because you cannot use the sort key “by bootstrap” data sets when doing simulations. The reason is that each data set has its own set of population characteristics and therefore the sort key cannot be used as only one set of population characteristics is available during simulation procedure.
The only way to do it is to split your data set by bootstrap and then simulate from each individual bootstrap data set.
I wish there was a better solution but you understand the reason beyond that option.
best
Serge
Hi Raghav, Hi Serge,
A common scenario when we need to bootstrap by replicate is we are conducting a Power Analysis study where for each replicate our decision depend on the confidence intervals generated via bootstrap.
An example of a decision dependent on Bootstrap for an FDA related work is for Crossover Pharmacodynamic Bioequivalence trials where success and failure is based on whether you bootstrap CI is within the decision limit.
The workflow to Power the study is like the followign:
Simulate the design for the trial using the baseline number of subjects N= 24 total subjects. Repeat the simulation enough number of times to reach convergence ( means results do not depend anymore on the number of simulation performed) and also for each simulated trial we need to Bootstrap it to ( again enough number of times till the CI are not dependent on the number of Bootstrap replicates performed) and then see whether it was sucessful or not and so on. This will give you power for N= 24 and then we need to go up or down depending on the results untill we reach 80-90% Power.
This is currently too tedious/undoable in Phoenix GUI I recommend users to try to use command line and use scripting for these kind of scenarios.
then incremently increase or decrease N ot detect the scenario that has 80 or 90 % Power for yeae
Hi serge,
I have a data set of 10 subjects, and i wanted to replicate to 1000 data sets . Is there any way we can do that either using phoenix or not using phoenix ?
Regards,
Raghav.
