1

In a randomised control trial, true randomisation is applied to decide which subjects receive treatment. In quasi experimental trial, the randomisation is only approximate.

What is the name given to such a trial where subjects are both treated and not treated at the same time. I am conducting "digital" trials, where I am able to test a "treatment" (obviously not directly on humans) and no treatment simultaneously under the exact same conditions. I believe that, theoretically, this should be superior to a traditional experiment, due to (synthetic) identical conditions.

What is the name for this type of experimental design?

Gilly
  • 247
  • 3
  • 8
  • 1
    I am not sure what you are talking about. Do you have multiple units of randomization per subject (eg two eyes in a man or four wheels in a car)? Then it's going to be a randomization with clustering. Quasi applies to non-random allocation, as far as I am aware (eg date of birth, which may be correlated with some features). – Giuseppe Biondi-Zoccai Mar 22 '16 at 11:07
  • No, no randomisation at all. I want to test the effect of `A` vs. `B`, so I apply `A` to all subjects, and apply `B` to all as well, then compare the results. This should be valid for deterministic systems, like algorithms. So I suppose its not a trial at all. But it still is a study - what do you call it? – Gilly Mar 22 '16 at 11:15
  • 1
    Then it's simply a cross-over/before-after trial. But you could still randomize the sequence. For instance, half of the subjects will receive A and then B, and the other half B and then A. – Giuseppe Biondi-Zoccai Mar 22 '16 at 11:17
  • I'm not sure. A before/after trial implies that the treatments are applied at different times with one (possibly) affecting the results of the other. I didn't mention specifically to keep the question as general as possible, but my project relates to applying different image processing methods to diagnostic images retrospectively. Each method can be applied completely independently. – Gilly Mar 22 '16 at 11:38
  • Maybe I can't find a name because it does not really count as a study? I'm not so sure anymore :S – Gilly Mar 22 '16 at 11:39
  • If sequence does not matter and analyses are completely independent then it's simply a comparison of paired data. Randomization is a feature relating to the underlying process. If such process is meaningless, then there is no randomized label to attach. – Giuseppe Biondi-Zoccai Mar 22 '16 at 11:47
  • Let us [continue this discussion in chat](http://chat.stackexchange.com/rooms/37333/discussion-between-gilly-and-giuseppe-biondi-zoccai). – Gilly Mar 22 '16 at 12:01

1 Answers1

2

A randomized trial allocates units of study (eg subjects or schools) to different approaches (eg treatment A vs B) using a random process (eg random number tables).

You may do a randomized cross-over trial using the same units of study, in which case the sequence of approaches is randomly chosen (eg some subjects will receive intervention C followed by D, and others intervention D followed by C).

A quasi-randomized trial does not exploit random allocation (but quasi-random approaches, such as date of birth or day of the week), and thus leads to potential bias.

If randomization is not pertinent to your project, eg Google using two different ordering algorithms on the same data, then you are simply performing a (experimental) study using paired data.

Giuseppe Biondi-Zoccai
  • 2,244
  • 3
  • 19
  • 48
  • Thank you very much. In my case, the last option is the most likely candidate: no randomisation so an (experimental) study using paired data – Gilly Mar 22 '16 at 12:52