Lurking Variable Question

[mjdav]

These question drive me crazy.

I guessed incorrectly twice, both C and A. For some arbitrary reason D is correct. Why? The question says clearly that the nurses filled out a questionnaire "about their health and habits." So in what way does "health consciousness" NOT fall under "habits?" As for that matter would being in shape (how much they exercise, if at all), no?

I guessed Family History because while the nurses answered questions, they themselves may not have known this information about themselves, and asking about your health and habits would not to my mind lead to asking about family history automatically. The study did not look at family history and we're told that's one of the requirements for a lurking variable.

This question and those like it are why I hate Statistics. I use all of my reasoning ability and I'm wrong repeatedly, and I'm absolutely unable to see the distinction here. I would appreciate anyone dumbing this down for me.

 

[Dr.Peterson]

This question and those like it are why I hate Statistics. I use all of my reasoning ability and I'm wrong repeatedly, and I'm absolutely unable to see the distinction here. I would appreciate anyone dumbing this down for me.

Questions like this are why I don't like statistics being taught with computer-based assignments that don't allow for the fact that this is not mathematics! This sort of question doesn't have one clearly right answer, as most math problems do; it depends on your judgment of non-mathematical factors. It's a great question for class discussion, but is only frustrating in this setting. (I've worked with a lot of students who struggle with this just as you do, and I admit to them that I struggle too.) Don't let it bother you, and don't let it sour you on actual statistics that you will be learning. Statistics happens to tie in to real life better than many kind of math, which opens it to this sort of question, but also makes it far more interesting if you care about real life.

I would recommend asking your teacher about it, so you can discuss the issues. I haven't taught the course, so I can't be sure what they have in mind; it's just possible that your teacher would agree that it is unclear, or there may be something taught in the book that makes it clear to them.

My best guess is that a lurking variable is something they are not taking into account that might be correlated with taking the drug, and being health-conscious fits that. It is not an explicit habit, or an explicit aspect of their health (as being in better shape would be), and it is not unrelated to choosing to take a drug (as family history may be). But this is not, to me, something that should be obvious to you if you understand statistics enough.

But someone who does teach statistics may have a better idea. And, again, your book may have said something that would more clearly distinguish lurking variables from other things.

 

[mjdav]

That is actually a great explanation.The lurking variable definition in the book is something not looked at in the study and that could be interacting with both variables. I can see family history not fitting this because while it might lower your risk of heart disease, your decision to take a pill or not would either be unaffected (if you didn't know the family history) or you may be if anything less inclined to take the pill.

As for health-consciousness. I also can see what you're saying about it not being explicit. But in a classroom setting I would definitely argue that health consciousness, at least to me, would manifest as things like diet, exercise, seeking medical care, etc. and that those things would be obvious - if your sample is full of health nuts, you would probably have that information given their line of questioning regarding health and habits. So I guess this all hinges on whether or not I think the authors thought the way you did or the way I did. Maybe I could get there by process of elimination as it's the best option available.

But really I'm glad to see your point about this not being math. Because the scenario above and whether or not health-consciousness was explicitly looked-at and the implications of all of this seems way too complicated for a multiple-choice question that you need a scalpel to dissect, along with a fair amount of trying to mind-read the authors. Unfortunately my class is all online and there's no real interaction, no elaboration, and posted questions don't get answers that I find useful or clear. I have sent a couple of these questions to the teacher and he agreed, actually flagging one to the publisher that was terribly-worded and impossible to answer as it was phrased.

Below is another one. I knew, because the book states it explicitly, that all major surveys suffer from some degree of non-response bias, so it was A or C (A was correct). But how in the world is there sampling bias? I thought ok, maybe we're saying it was a mailed survey, but they don't say that, and what kind of doctor doesn't have an address? The population was all doctors, so I don't see the problem there with the sample.

I thought maybe response bias because maybe some of the doctors shill certain drugs vs. others or have a reason to want the price lower and tend to respond because they feel more passionately about it, but frankly I know almost nothing about doctors' relationships to prescription drugs.



Anyway, thanks for your explanation above and I will try not to take these personally. I think to really understand this subject I would need a live class with questions and discussion.

 

[Dr.Peterson]

How does your book define "response bias"? I have no special knowledge of it, but to me it's a matter for deep study of psychology, not something a one-paragraph explanation in an introductory chapter on sampling can teach adequately.

My main point (again, biased because I don't teach this but try to tutor students taking it!) is not so much that you need a live class to understand it (though that may be true), as that these particular questions don't necessarily have clear answers, so the best answer is a discussion that brings out various sides. In another course, one that dug more deeply into it, they would be essay questions. My usual hope is that they don't show up, at least in their worst form, on tests.