Potential Confounders in medical research.

[Fire_Medic_Dave]

Ok I know what a confounding variable is. My question is what role does a "potential" confounder play in a statistical model? It seems like an ambiguous term. Just wondered if any of you statisticians might be able to clear it up.

David

[chicagomd]

1. My understanding

is that when authors point out "potential" confounders it refers to something that may change the data set before statistical analysis but is unable to be, or was by study design not, controlled for. Essentially they are pointing out variables that need to be considered when you are looking at the statistical outcome, as they are not considered within the model used.

For example, a potential confounder in a research phone poll on who people are going to vote for in November is prejudice against African Americans. People might say on the phone that race is not a variable, but it potentially is not measurable because you cannot control for people's level of dishonesty.

It is also a nice way of pointing out the limitations of your own study and beating the critics to the punch.

[Fire_Medic_Dave]

2. Thanks for the explanation.

Seems to invalidate or significantly weaken conclusions made from studies where confounders are not factored into the model. Especially if the confounders are statistically significant.

David

[WildClarySage]

3. But maybe this needs to be posted a few more times

before everyone is convinced. I mean, repeated posts on the same topic *must* be more truthy.

[Fire_Medic_Dave]

4. Sorry I've never seen this posted, I was just trying to gain some knowledge.

[WildClarySage]

8. I get that. I was referring to the numerous

repeated threads started by the same person saying exactly the same thing in each thread.

[Fire_Medic_Dave]

9. No worries.

Didn't know where you were going with it.

David

[cosmik debris]

5. I believe that a "potential confounder"

could be used to introduce other sources of error.

If you are trying to correct for a "potential confounder" you have to assign a numerical value for the "potential" of occurrence. And you also have to have some numerical value for the impact of the "confounder".

Any variability in these two values will perturb your statistical model.

But that is based on my limited understanding of the term.

[Fire_Medic_Dave]

6. That would make sense.

Thanks.

David

[xchrom]

7. kick