1
$\begingroup$

I have a dataset of an experiment that had the following design:

  • N=80 (40 male, 40 female)
  • 2 treatments ("control treatment" (CT) and "sensitized treatment" (ST))
  • Each treatment consisted of 20 males and 20 females
  • All subjects had to answer a question considering their risk preference (ordinal scale)

Now I have the risk preference values for each of the four cells: 1. males in CT, 2. females in CT, 3. males in ST, 4. females in ST.

I want to test if there is a difference between genders in terms of the effect of sensitization on the average value of the stated risk preferences. For example: Women were more prone to be affected of the sensitization and thus had a higher/lower average stated risk preference.

How do I do that?

Thanks in advance!

$\endgroup$
5
  • $\begingroup$ Have you thought of looking for an interaction effect? $\endgroup$
    – mdewey
    Commented Jul 26, 2018 at 12:53
  • $\begingroup$ in addition to mean, would you also have respective standard error or standard deviation within the cell? $\endgroup$
    – QmmmmLiu
    Commented Jul 26, 2018 at 13:42
  • $\begingroup$ @QmmmmLiu Sorry, of course I have all values for all subjects. Not just the mean values! $\endgroup$
    – tanja94
    Commented Jul 26, 2018 at 18:12
  • $\begingroup$ @mdewey I don't know how to do that. $\endgroup$
    – tanja94
    Commented Jul 26, 2018 at 18:13
  • $\begingroup$ If you search for stats at ucla interaction test and add in your favourite statistical software you should find pages from that university (UCLA) which explain with examples. $\endgroup$
    – mdewey
    Commented Jul 27, 2018 at 8:00

1 Answer 1

0
$\begingroup$

If you only have the mean for each of the four cells, then, as far as I know, you can't do any statistical test. You can see if the differences are additive or not by inspection, but you can't test that difference.

Usually, you will have the data on each of the subjects (here, the 80 people) and can run regression with an interaction effect (or ANOVA, it's the same thing). Of course, you will have to check all the assumptions of the model.

$\endgroup$
1
  • $\begingroup$ Sorry, of course I have all values for all subjects. Not just the mean values! That is, to show that genders are differently affected by the treatment, I should do a "regression with an interaction effect"? I don't know what that is, but at least I now know what to look for! Big thanks. $\endgroup$
    – tanja94
    Commented Jul 26, 2018 at 18:12

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.