sample size calculations and power

Question 1

What is the name of the graph which can be used to determine sample size or power?

Answer 1

Nomogram

Question 2

What needs to be specified to calculate the sample size for 2 independent samples of continuous data?

Answer 2

the clinical relevant difference
the standard deviation of the variable (in each group)
the significance level (two sided)
the power, 1-B

Question 3

What is the standardised difference?

Answer 3

the clinical relevant difference over the standard deviation

Question 4

What can be done when the standard deviation is not known?

Answer 4

from previous studies
start the trial then calculate SD (not ethical)
reconsider variables as categorical
Alternatively express power calculation in terms of standardises differences so that the effect detected is expressed as a function of the unknown standard deviation

Question 5

What standard deviation is needed for paired samples of continuous data?

Answer 5

the standard deviation of the change of measurement

Question 6

What is the standardised difference for paired samples?

Answer 6

2 clinically relevant difference / Sd

Question 7

What are the calculations of sample size for 2 independent variables of categorical data based on?

Answer 7

the expected proportion in each group with the specified outcome p1 and p2
the significance level (2 sided)
the power

Question 8

What can be done when the sample sizes are unequal?

Answer 8

so long as the imbalance is not great the loss in power is small
if k = n1/n2
then the total sample size is N*=N(1+k)2 / 4K