In a simple statistical analysis involving a control group and an experiment group, the research must demonstrate that before the experiment, the control group and the experiment group are the same or at least not significantly different. What is the name of such tests?
$\begingroup$
$\endgroup$
3
-
2$\begingroup$ The word that describes this is probably representative, although I wouldn't recommend using a significance test to assert this, or really the validity of any assumption. All tests have their own chance of a false positive and at best you'd find not enough evidence against the control being representative. $\endgroup$– Frans RodenburgCommented Oct 10, 2019 at 23:02
-
2$\begingroup$ Ideally, such initial comparability of the two groups is assured by random assignments of subjects from a common pool to the two groups. Alternatively, maybe the common pool is stratified in some relevant way (possibly: age, gender, ethnicity, etc.) into a few subsets, which are then randomly assigned to the two groups. Maybe best just to say briefly what you did to ensure that the two groups started out as comparable. $\endgroup$– BruceETCommented Oct 11, 2019 at 0:21
-
$\begingroup$ @BruceET In a natural experiment or a quasi-experiment the researcher might not have the chance to assure random assignment -- so the research has to make additional efforts to "prove" that the control and the treatment group are indeed similar. $\endgroup$– High GPACommented Oct 11, 2019 at 21:07
Add a comment
|
