Confounding vs. Extraneous Variables?

I need to know the difference between these two in a psychological research paper I am studying, in which there are two levels of independent variable. The first level is a group with reads in silence, the second group reads aloud. However, there is 48 males and only 26 females in the entire sample. The procedure does not indicate that there was random allocation of groups. Would this mean that we could suspect gender as a potential confounding variable? As there is no evidence that one of the groups was not a group made up of entirely males! However, a confounding variable needs to systematically vary across the levels of independent variable, and correlate with the dependent variable, but I cannot tell if it does or not! Is gender a confounding variable here?


No cake for spunky
It is nearly impossible to answer this question without seeing the article. Too many details are missing. If gender is not in the model, if there is some reason to believe theoretically that it might influence the DV (or alternately be correlated with both the IV and DV) and you don't have random assignment than gender could be a confound. Extraneous can mean many things, most commonly it means a variable not included in a model and it is hard to say what is meant by it without seeing the article.


Less is more. Stay pure. Stay poor.
Agree with noetsi, more information needed. So there are two predicting variables in the model? They are both significant? The authors did not look at any additivie or multiplicative interaction terms? What do the authors state in the discussion section of the paper?


No cake for spunky
The term confound is usually used in the context of design of experiment and means that you can not seperate out the effect of one variable on the DV from another. Without random assignment this is virtually always true. In statistics you do not really address confounding, a theoretical concept. Instead you focus on the unique variance explained in the DV by one variable through statistical controls. Part of the variance in the DV will be explained by both variables, which is arguably confounding, unless the IV are not correlated at all which is I suspect very unlikely.