Normally Distributed Data Sampling and T-test Simulator
FREE on-line and off-line tools
(last updated on 2019-11-12)
The tool (on-line version requires IE) simulates the sampling and t-test of two groups of independent,
normally distributed data such as the treatment and the control arms of a clinical
study. To carry out the simulation, the mean and the standard deviation of each
group, which are unknown normally, need to be hypothetically specified. The simulation
generates random numbers as samples based on the specified mean and standard deviation.
It shows the relative distribution of the sampled data in the charts above and below
the button Simulate, then it computes the sample statistics – mean, standard deviation,
t value of Student’s t-test, one-tailed and two-tailed confidence levels.
The simulator can help provide some sense about the outcome of a study under different
scenarios (e.g. different standard deviations, sampling sizes). It can supplement
power analysis. For
example, suppose a study on a cholesterol lowering drug is about to be conducted
and one measure of the outcome is the cholesterol level change in one month.
In addition, it is hypothetically assumed that the mean and the standard deviation
are -15 mg/dL and 20 mg/dL respectively for the medication arm. The corresponding
parameters for the control arm are 0 mg/dL and 10 mg/dL. Different sample
sizes can be entered to see how they affect the statistical significance of the
difference between the treatment arm and the control arm. The standard deviations
and the means can also be altered to see their effects. The same set of parameters
can be run multiple times to see how the effect of the inherent random nature.
Please note: Due to the random nature, every run of simulation generates a different
set of results though the distributions of the sampled data are governed by the
specified means and standard deviations.