This is a graphic representation of the Time Series Design:

You make a few observations to establish a baseline, do the intervention, and then make a few more measurements. (Just how many observations you make on either side of the intervention is determined by the stability of the data. This will be covered in advanced concepts.)

Time series quasi-experiments can have any number of outcomes. They can be graphed as follows (some examples):

The interpretation of these outcome graphs is simple:

In 'A', the outcome measure is stable at a certain value until your intervention, then it increases and stabilizes at a higher level. In our diabetes example, this would be equivalent to having the intervention "backfire" (i.e., blood sugar levels increase instead of decrease).

In 'B', the outcome measure is again stable at a certain value until the intervention, then it decreases and stabilizes at a lower level. In the diabetes example, this would be equivalent to success (i.e., blood sugar levels become consistently lower after the intervention).

In 'C', the outcome measure is stable at a certain level until the intervention, then it increases immediately after the intervention; however, it then returns to the pre-intervention baseline. Something may have happened as a result of the intervention (e.g., an increase in blood sugar level in the diabetes example), but it was not permanent.

In 'D', the outcome measure is - well - basically all over the place. In a situation like this it will be tough to "tease out" the effects of your intervention. There may be a stable effect of the intervention in there somewhere, but it is not readily obvious. This is an example of a situation where you would have to give some thought to how many pre and post observations would be needed to gain enough statistical power to detect a result (if it is there).

The major threat to the internal validity of the time series is history. That is, a charge that the results obtained would have occured with or without the experimental intervention is difficult to defend with data from the simple Time Series experiment.


Return to quasi-experimental designs.