Hello,
I'm working on a study that consists in evaluating to what extent an event impacts the perception of a brand.
Basically, the idea is to evaluate this perception, to present the respondents with an event, and then to see to what extent this perception evolves.
2 possibilities:
- I do a within subjects study: we keep the same people between the two questionnaires (before & after the event)... but as we have to wait one or 2 weeks between the 2 steps, I can't know how many people should be interviewed in the first one (I don't know how many will be there in the 2nd round)
- A between subject study. This allows me to reduce the number of interviewees, to save time, to control the number of respondents & therefore to reduce the collection costs. The idea would be to randomly assign each respondent to one of the 2 situations; without the event in question, or after this event. If the 2 populations are identical in terms of profile, this would allow a comparison.
I have never seen a study of this type done in a between subjects setting, do you have any references? If it has already been done!
Thank you.
ps: Sorry for the English mistakes
A within-groups design is the better choice here. You will require fewer participants and have more statistical power. How many participants you include in the sample depends on the anticipated effect size. I recommend conducting an a priori power analysis with various effect sizes to get an idea of how many participants you should include.
Thanks Blaine Tomkins !
i also think like you but some colleagues recommended a between-groups deign without really being able to explain why (except for the reasons I gave).
I've strangely never done any power analysis, could you please provide me link to find out how to do that?
Thanks in advance,
mk
Michaël Korchia The only potential risk of using a within-groups design is the possibility of tipping participants off regarding the goal of the study/hypothesis - referred to as a demand characteristic. For instance, if you measure my attitude about a brand once, present me with a manipulation, and then measure my attitude once more. I'm likely to guess that the purpose of the study is to see if my attitude changes - which may influence the way I respond on the posttest. A between-groups design eliminates this risk.
Regarding the power analysis, I recommend using G*Power, a free software available here:
https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower
Conducting a power analysis is fairly straightforward in G*Power. Take a look at the available tutorials to learn how to use it.
Blaine Tomkins , sorry for being late at replying, I've just seen the notification in my mailbox :/
Thanks a lot! Very helpful. I'll do what you said.
David L Post
thanks!I also found another possibility, in an article. A post test only study.
The control group would allow us yo make a comparison. Probably the most simple/cheapest way to go!
Michaël Korchia -
A post test only? Not my area. Sounds promising, but also sounds like you are going to have to make some assumptions about the control group. - Is the article open access? If so, maybe Blaine and/or David and/or someone else may have a comment or two on it.
Cheers - Jim
- Badges
- Science topic
- Similar topics
- Linguistics
- Philology
- Language
- English