1

I am reviewing a number of studies for my thesis that all manipulate the same factors A (with levels A1 and A2) and B (with levels B1 and B2). Now some of them only manipulate A while keeping the level of B fixed (comparing A1B1 to A2B1, for example), and I can refer to them as A studies (analogous with B). However, some unwittingly manipulate both A and B at the same time and compare A1B1 to A2B2.

I am not discounting those studies, as they are the vast majority and as such nicely illustrate the problem with the field. But I also use them to point out areas worthy of study (If 10 studies find a difference between A1B1 and A2B2, then there must be something there, we just don't know if it's due to A or B or both). So I don't want to refer to them as confounded studies, which is what they technically are, but which I find to have too much of a negative connotation.

I am calling them crossed studies at the moment, but am afraid that will be confused with a fully crossed study, which is the accepted term for comparing all possible combinations of A and B, and which is decidedly what those studies don't do.

  • 1
    This is probably too specific to expect that a single word exists to describe it. I'd assume that 'trivariate' is confined to functions of the form z = f(x,y), and studies addressing these directly rather than by chance. – Edwin Ashworth Jun 16 '15 at 10:26
  • @EdwinAshworth trivariate would make a statement about the number of variables, but that number is the same for all types of studies. I was hoping there is an accepted term in scientific discourse that I am simply unaware of. – ThomasH Jun 16 '15 at 10:39
  • @ThomasH I'm with Edwin Ashworth. It's very doubtful that a single word exists that describes what you're asking for. I'm certainly not aware of one, and I'm an analyst. Why not ask your thesis adviser? At the end of the day, it's possible to overthink some of the truly trivial details about your thesis. Why use a single word when an overblown paragraph will do? – DJohnson Jun 16 '15 at 10:41
  • @MikeHunter I'm prepared to accept that there isn't a term that fits my needs. Unfortunately, I have about 70 pages in which I constantly refer to A studies, B studies and possibly studies that change A and B at the same time, the latter of which becomes rather unwieldy very quickly. So I might have to stick with crossed studies for now and just add an overblown paragraph at the start to explain my use of the term. – ThomasH Jun 16 '15 at 10:45
  • 1
    BTW, what is the domain? – Kris Jun 16 '15 at 11:50
1

Why not qualify "crossed" as "partially-crossed" or "quasi-crossed"?

1

A few thoughts about this:

1) Personally, I think it is worth making scientific writing as plain and accurate as possible. If "confounded" indeed describes the relationship between A and B, then that is the term you should use.

2) I don't think "confounded" has a negative connotation at all* - at least in the epidemiology literature. It merely specifies a specific relationship between variables (which you already know.) I'll use the ubiquitous example: in the apparent correlation between alcohol use and heart disease, alcohol use is confounded by smoking status. Not positive, not negative, just descriptive information.

(*Although there's "that paper confounded me!" Or, "that #^~& confounded paper!" But these are different.)

3) Now, you are considering using the term slightly differently, applying it to a whole study rather than variables. But I do not think this shift adds a negative connotation.

Some degree of confounding is ubiquitous; the real issue is the quality of the study. A lousy study can be useless even if there are no major confounders. On the other hand, even if substantial confounding is unavoidable, thoughtful design, skilled implementation, and shrewd analysis can still provide valuable information. So - the presence of confounding alone does not imply any deficiency in the research; nor does it prevent useful study results; I don't perceive any negative connotation.

4) However, my experience is in epidemiology, biostatistics, and medicine - which may or may not be relevant to your thesis. It's difficult to say, without knowing the topic or even the field. Or you may just disagree. So I have another suggestion, in case you decide not to go with "confounded."

A common way to deal with this situation is assign each group an abbreviated or otherwise convenient name, define the names early on (e.g. in the introduction and/or methods section) and use the short names thereafter. Some possibilities: -Group 1, 2, and 3 -Group A, Group B, and Group AB -Something more descriptive (e.g. AVS - A variable studies, BVS - B variable studies, and ABVS - A and B variable studies. Personally, I prefer the descriptive abbreviations; otherwise, I too often forget which group is which.)

5) Finally, I agree with checking with your thesis advisor or committee. They may have very specific requirements or preferences. And good luck!

0

Could they be called uncontrolled?

My reasoning is that if B is kept constant then A1 can be thought of as the control group of A2 (expanding the meaning somewhat but not much) and vice versa but if B isn't then they can't.

0

Personally I would put a table in the intro that lists the variations with a name for each. Either a simple logical progression (group 1, group 2) or something more descriptive if you prefer - then go ahead and use those definitions throughout.

-2

Can "Simultaneous equations" be tried??----a suggestion only from one novice at that!!

  • "Simultaneous equations" is a method for solving equations that have more than one unknown. How does that apply here? – Matt E. Эллен Jul 6 '15 at 13:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.