The data for this study have already been loaded into this applet. Verify the two dotplots and the observed difference in group means agree with our data.
Then check the Show Shuffle Options box.
Select the Plot display.
Then press Shuffle Responses button once and watch the dots. The group assignment (color) of the dots were be randomly determined and then the values will be redistributed back to the newly assigned groups.
Once the improvement scores have been randomly redistributed to the groups, the group means have changed and so has the difference we compute between the two groups (unrestricted - deprived). This new could-have-been difference in group means is added to the dotplot to the right (scroll to the right).
Press the Shuffle Responses button again. Do you get the same difference in group means this time?
Change the Number of Shuffles from 1 to 18. Press the Shuffle Responses button. Are you beginning to see a pattern in the bottom dotplot?
This is a good time to ask if you don't understand what is going on here. (Hint: What are the observational units and variable in this dotplot?)
(scroll down for more instructions)
Change the Number of Shuffles to 980 and press the Shuffle Responses button.
To approximate the p-value, we want to know how many of these simulated differences in group means are at least as large (in the direction conjectured by the researchers) as the one observed. Recall the value of the difference in sample means observed in our study (see (d)), and type that in the Count Samples box depending on whether you want to count greater than, less than or beyond (both sides). Then press the corresponding Count button. The dotplot should shade the values at least as extreme and report the empirical (estimated) p-value.