1
$\begingroup$

Is there a use of using a Randomized Block design when you only have one treatment and you only want to test the effects of this single treatment?

The experiment in question is to see whether a training exercise (the treatment) effects the grades of a number of students. Students are grouped in a control and experimental group and are subject to a test. The experimental group are then given training and then asked to re take the exam.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

This'd be a block design, but decidedly not a randomized one because the subjects receive the two treatments (i.e., treatment levels) in the same order. This kind of design is often called a pre-post design. Analysis is often done using the gain scores -- post minus pre.

Improve this answer
$\endgroup$
4
  • $\begingroup$ PS I'm assuming that the control subjects are also re-tested! Otherwise you have no way of sorting out the effect of retaking the test from the effect of the treatment. $\endgroup$
    – Russ Lenth
    Commented Aug 30, 2014 at 18:10
  • $\begingroup$ Yes control subjects were re-tested. I still cant understand how this is a block design. $\endgroup$
    – user54967
    Commented Aug 30, 2014 at 18:16
  • $\begingroup$ In this case the experiment was done in 5 different centers and 60 pupils were randomly subdivided into control and experimental groups (30 each). Would this be considered a block design? or would have the the 60 pupils be required to be grouped in some way before randomly selecting them and assigning to the experimental/control groups? $\endgroup$
    – user54967
    Commented Aug 30, 2014 at 18:23
  • $\begingroup$ Indeed, some people would not call this a block design -- maybe even me -- because technically a block design has a restricted randomization on each experimental unit. But what's important is not what you call it, as long as you do a sensible analysis. If you're using the gain scores, then what you are left with is two samples of 30 gain scores and you want to compare their means. Which you could do using an "independent samples" t test, pooled or not. But you mentioned different centers -- and wonder if we need to take their effects into account. That'd be a different model, and not an RCB. $\endgroup$
    – Russ Lenth
    Commented Aug 30, 2014 at 18:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.