# Thread: What is the difference between an interaction effect and a moderator in regression?

1. ## Re: What is the difference between an interaction effect and a moderator in regressio

Perhaps you need to find some examples with data, and see if these things hold. If you had a continuousXcontinuous interaction, you could always bin one of the variables into groups and play around with some of these concepts.

2. ## Re: What is the difference between an interaction effect and a moderator in regressio

Or so I was told in my program. I am still not sure how SAS handles interaction diagnostics or simple effect with two interval variables. Perhaps there is no reasonable way.

3. ## Re: What is the difference between an interaction effect and a moderator in regressio

It is not bad bad bad,..., when you are just playing around with concepts for fun. I get the loss of information, but you are more likely to find a practice dataset with two continuous variables than interval variables.

4. ## Re: What is the difference between an interaction effect and a moderator in regressio

but you are more likely to find a practice dataset with two continuous variables than interval variables.
I don't think it is possible to find a practice data set with two continuous variables that are not interval

I think you meant categorical rather than continuous.

5. ## Re: What is the difference between an interaction effect and a moderator in regressio

Or there are differences in data across fields. For example I rarely come across interval data.

6. ## Re: What is the difference between an interaction effect and a moderator in regressio

That's amazing to me. I commonly work with money which is entirely interval.

7. ## Re: What is the difference between an interaction effect and a moderator in regressio

Oh my bad. I rarely use these "types" for classifying data (to me continuous, discrete, and categorical tend to be "good enough"). For some reason I read that as "ordinal" instead of "interval".

8. ## Re: What is the difference between an interaction effect and a moderator in regressio

Is discrete data count data? When data is a count, in practice, is rarely obvious to me. For example I never run Poisson regression with our data although arguably some of our results are counts [like how many times you receive services].

I think of data usually as interval, ordinal, and nominal the last two being important distinctions when I run logistic regression. We have lots of ordinal data when we administer likert type survey data [although commonly we treat that as interval....]

9. ## Re: What is the difference between an interaction effect and a moderator in regressio

Discrete data doesn't necessarily have to be count data. Gender is also discrete, i.e. you can only be male or female (ignoring politically correct arguments here..).

10. ## Re: What is the difference between an interaction effect and a moderator in regressio

Back to original topic:

On the Distinction Between Interaction and

Effect Modification

Tyler J. VanderWeele
Epidemiology • Volume 20, Number 6, November 2009, pg 863.

Feel free to read the whole thing and report back