Internet
Demonstrations
Up
| | You can link to an experiments applet
to simulate outcomes for either
two independent or two matched samples of varying sizes, given specified
differences between populations.
Two independent samples are referred to as between-subjects designs, and
two-matched samples are referred to as within-subjects designs.
If the two populations are independent, the correlation coefficient (rho)
between populations equals zero. If, however, the two populations are dependent
– and therefore a two-matched or within-subjects design is preferred – the
degree of dependency is specified by a non-zero value of rho.
The formulas for t differ from those in the book, but the resulting value of
t is exactly the same for both versions.
- Read Instructions.
- Use the initial values in the demonstration, and at least 5,000
simulations to estimate the proportion of correct (statistically
significant) outcomes, first for the within-subjects design, than for the
between-subjects design. (To make this comparison, simply pressing the reset
button won't permanently reset the cumulative total of significant and
nonsignificant outcomes. Instead, once you've completed the within-subjects
simulation, exit from the applet window, then begin again, now specifying a
between-subjects design.) You’ll note that, because of the rho of .50
between populations, the within-subject design leads to appreciably more
correct (significant) outcomes than does the between-subjects design.
- Click on "Begin" to start demonstration, and when finished, click on
"Exercises."
To check your responses, click on "Explanation."
|
Experiments