Confounding and extraneous variables


New Member
Hi all,

Can someone please clarify the difference between a confounding and extraneous variable? I've read tonnes of explanations but still can't get my head around it.

So far i've come to the conclusion that a confounding variable is a type of extraneous variable that varies systematically with levels of the IV. Fair enough, but how does an extraneous variable affect the DV then?

Lets say we are using emotion regulation skill (IV) to predict depression (DV), but we know that gender is related to both the IV and DV so want to control for it. Does this make gender a control, extraneous or confounding variable?

And what if we also discover that education level is related to emotion regulation (the IV) but not depression (the DV) when we didn't expect it to be related to either? I thought this would be a confounder. Is this right?

Any help greatly appreciated.
I'm in Australia, and we have really only used the term confounding variable. According to my text book (Statistics and Research Methods; Davis & Smith), confounding relates to the IV, and extraneous to the DV. Its all a bit arbitrary in my opinion. If the IV is affected, then so are your results. If the DV is affected, then so are your results. The confound comes in when something other than the IV can explain or contributes to your data.

In your example, there are heaps of different things that could explain depression besides emotion regulation skill, like gender, education level, geographic location, income, employment, genetics, childhood experience.... and all will confound your study to some degree. The best thing to do would be to select a population that are similar in these constructs, so as to control for them, and then state in your study that your research only extends to that given population. Anything you can't control for should be mentioned in your study.
I'm trying to get a better general understanding of blocking and confounding in the 2^k factorial design. So far, I already understand about confounding interactions with blocks when it comes to picking one replicate in a data as you can see in the top table of that Excel file uploaded here. That's where I confound ABD and ABC (and consequently CD) with the 4 blocks assumingly required.

As you can see from the bottom table of that Excel file, what if I was asked to confound ABCD in one replicate (let's say replicate I) and ABC in the other (replicate II)? Do I use the same approach like I would for just one replicate?