additive versus multiplicative

[noetsi]

a high level book I am reading says there is a difference between additive and multiplicative interaction. And they can, for example generate different signs for the same data. What are these? I never heard of different types of interaction.

 

[noetsi]

Using this definition I don't think I have ever seen additive interaction discussed before.

"The presence and direction of interaction depends on the scale, e.g. additive or multiplicative, that is used. Interaction on an additive scale means that the combined effect of two exposures is larger (or smaller) than the sum of the individual effects of the two exposures, whereas interaction on a multiplicative scale means that the combined effect is larger (or smaller) than the product of the individual effects."

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3115067/

 

[hlsmith]

Correct, the very simple way to look at it was to calculate the outcome metric, say the risk, for exposure one and exposure two then see if when both are present that metric is greater than their presumed individual effects.

 

[noetsi]

I can't think of how that would occur in the social security world I do analysis. What would an exposure even mean in that case?

 

[hlsmith]

A general example I use is smoking and asbestos exposure on lung cancer. Other examples are environmental and genetic on diseases.

Any place where there can be a multiplicative interaction, that is the setting where additive can occur. People just use multiplicative as the default since it is so easy to do. If one had theory, they would then be able to speculate on the scale for the interaction. I feel like I am now able to see more interactions in social contents or decisions. For example, I am working on a study where at some point in maturation, kids do something intentional that they only did accidently when younger, also a different study where decisions are different based on perceived societal constructs.

Think about age and decisions, a current study I am working on older people are more likely to have vaccine acceptance and more educated people as well. So, I need to think about whether older more educated people will have an even greater acceptance than either of those individual groups. Of note, I think it is Donald Spiegel (sp?), that makes statements he has only seen a few legit interaction in his storied career in analytics.

 

[noetsi]

I never encountered the concept before reading "Explanation in Causal Inference by VanderWeele of which I have understood less than one percent. He talks about the same type of stuff you do I was hoping to use it to understand interaction and mediation but much of it is beyond me.

 

[hlsmith]

Keep reading, it'll start making sense. Post any questions, I have the book somewhere in my basement.

 

[noetsi]

Keep reading, it'll start making sense. Post any questions, I have the book somewhere in my basement.

Which reflects just how much you are beyond me.

 

[hlsmith]

Nah, I am sure you know a lot more time series than I do, its is just domain specific. That and I attended a 4-hour workshop in 2012 on interaction and a 4 hour workshop in 2014 on mediation both presented by Tyler.

 

[noetsi]

hlsmith one thing is for sure. Your understanding of complex statistics is way way beyond me. Same for Jake and Dason.