Summarizes the criteria for the assessment of randomization procedures.
Details
Randomization in clinical trials is supposed to
control certain properties in clinical trials. In the randomizeR package,
these properties are called issues.
It is crucial to decide which of the issues is relevant in the present
clinical trial, because a randomization procedure that mananges well one issue
might behave very badly for another. The issues include
Selection bias
can occur if future treatment allocations are predictable due to
restricted randomization and unmasking of past treatment assigments.
The influence of selection bias on the test decision is represented by
the selBias class. The measure for the predictability of
a randomization procedure is impemented in the corGuess class
representing the expected number of correct guesses.
Chronological bias
can occur if a time trend is present in the data. Time trends occur
due to learning curves, relaxed inclusion/ exclusion criteria or
new co-medication.
Chronological bias is represented by the chronBias class.
Additive combination of chronological and selection bias
may occur if a time trend and selection bias are present in the data.
The combined bias is represented by the combineBias class.
Balance
is important in order to ensure proper power estimation properties of
the treatments.
However, a high degree of balance favours selection bias. Depending on the
clinical context, a randomization procedure should be chosen that admits
a suitable imbalance.
Imbalance bias is represented by the imbal class. The power
loss due to imbalance can be assessed directly via the setPower
class
See Also
Representation of randomization procedures: randPar
Generation of randomization sequences: genSeq
Assessment of randomization sequences: assess
Comparison of randomization sequences: compare
Other issues: chronBias,
combineBias, corGuess,
imbal, selBias,
setPower