# Mediator versus moderators - not sure which is more appropriate

I am undertaking a study which has the following aims:

The main aim of the current study is to investigate the effects of food messages about the healthiness of foods on actual food intake.

Additionally, a secondary aim is to discover the extent to which impulsivity/sensation seeking, taste preference and BMI may moderate/mediate the effects of food-related messages on food intake.

I am not sure which is more appropriate in this case to mediate or moderate. My DV is food consumption and IV's are food messages, impulsivity/sensation seeking, taste preference and BMI.

It is an experimental study which involves three treatment groups which will receive either a healthy message, unhealthy message or no message about the cookie. They will be asked to taste and rate and consume 'cookies'. The taste rating will also ask questions about taste preference. I will then provide a questionnaire which gathers info about impulsivity/sensation seeking and also BMI.

I am leaning towards mediation but am not sure. I believe that those in the healthy message group will consume more cookies based on prior research but I also deem that the other variables will have a influence as well on cookie consumption. Basically those who have high impulsivity, taste preference for the cookie and higher BMI either individually or in combination will consume more cookies.

So I am not sure as I've indicated whether it would be more suitable to have moderator's or mediators. I am somehow leaning towards mediation but would like some clarity around this.

Also if it's either mediation or moderation which IV's are which? Also what would a hypothesis look like?

I am guessing that if I went with moderation for example then a hypothesis may look something like:

the relation between health message and cookie consumption could be stronger for those who score higher on the impSS as compared to those who don’t score as high

But I may be wrong. Any help would be appreciated.

Thanks.

MO

-

This

the relation between health message and cookie consumption could be stronger for those who score higher on the impSS as compared to those who don’t score as high

although not in exactly the right language, is the central idea of moderation, you are right.

Mediation, on the other hand, means that the effect of one independent variable on the dependent variable is due (wholly or in large part) to a second independent variable. Here are three classic examples (the last one is both moderation and mediation):

1) Students who hire tutors get worse grades than students who do not hire tutors.

This is because the effect of getting a tutor is mediated by initial grades - students who are getting A's don't usually hire tutors

2) If more firemen show up at a fire, more damage is done. This is because the effect of firemen on damage is mediated by size of fire. Bigger fires get more firemen and more damage.

3) In early elementary school grades, there is a correlation between astrological sign and IQ. This disappears with age.

This is because the effect of astrological sign is mediated by time spent in school, and school systems have calendar cutoffs. So, a 6 year born in December could have been in school for 11 more months than one born in January. But this effect is moderated by age, because it becomes smaller as the number of years in school increases.

-
(+1), especially for the examples. –  Matt Krause Apr 13 '12 at 16:44
Hi Peter, thanks so much for your response. How do you then decide which is better either moderators or mediators? I believe it should be based on theory or what you think may happen in real life but are there rules that govern whether it is best to use one over the other ie. - type of study, type of variables etc? What would a hypothesis look like in my study that utilised mediation? Also how do you decide which is the predictor versus mediators/moderators? As my main aim is with health messages I assume that is the predictor. Sorry for so many questions. Any help would be appreciated. –  mobo Apr 13 '12 at 23:50
@Matt Krause Hi Matt, thanks for your comment. Please view my latest set of questions if you like as I'd be interested in your views also. Cheers mobo. –  mobo Apr 13 '12 at 23:56
@mobo You decide all this based on theory and what you think may happen and what you want to discover –  Peter Flom Apr 14 '12 at 11:46
I'm not sure "This is because the effect of getting a tutor is mediated by initial grades" is the right interpretation. See the schematic here: en.wikipedia.org/wiki/File:Mediation.jpg - for initial grades to be the mediator, you'd need, essentially ${\rm tutor} \rightarrow {\rm initial \ grades} \rightarrow {\rm final \ grades}$, which clearly isn't happening. Maybe you've interchanged the role of 'tutor' and 'initial grades' or I've misunderstood - in either case (1) sounds more like an example of confounding than mediation... There seems to be a similar possible error made in (2)... –  Macro Apr 15 '12 at 2:48