6

I currently am writing a paper which uses regressions (from statistics) heavily. The objective is to show (as far as possible) that A causes B.

The main analysis of my paper uses regressions to show that A is correlated with B.

After showing this correlation, I wish to show that A occurs before B occurs so as to give further evidence of a causal relationship. This requires another set of regressions. (The set of explanatory variables changes significantly, so I need to specify the new regression equation.)

Furthermore, I need to show that factors C, D, E mediate the relationship between A and B. (Again, I am much better off specifying the new regression equation. Even if you doubt this, just believe me on this)

How should I go about this?

Option 1: Explain all methods (correlation, temporal precedence, mediation) in the Methods section. Then give all results in the Results section.

Option 2: Explain the methodology for detecting correlation, then give the results. Then explain the methodology for temporal precedence, give the results. Then explain the methodology for mediation, and give the results.

The problem is, no matter which way I do this, I get criticism that the structure is confusing, along with suggestions that I should do the alternative approach. For example, if I try option 1, people say that temporal precedence and mediation only make sense if I detect a correlation, so these analyses should be shifted to the back. On the other hand, if I try option 2, people say that my methodology should all be neatly encompassed in a single Methods section.

I'm curious as to what most people do when they are in a similar situation?

4

Option 1 is probably a safer bet, because that's the logical structure of an article and it's there for a reason (to ease reading).

When faced with the same problem, I made a schematic representation of the analyses. I confronted this schema (and its improved versions) with colleagues that did not know the details of my analyses. This solution might not apply to you, but at least try. The old saying (an image is worth a thousand words) is applicable to scientific articles.

7

In my area the standard is always to have a single methods section; journals would not accept an alternative format. That might not always be the ideal arrangement for every paper, but it's simply the norm.

So what to do when the methodology is somewhat convoluted, or very key to particular elements of the results? Add some "reminder" description of the methods in the results. Keep the fine details only in the methods, but address what you are up to to keep a good story in the results. Minor repetition isn't a problem and helps your paper flow.

4

Begin telling the reader what you intend to do at a high level before diving into the details. This gives the reader a framework for how to think about your paper.

Then I would go with option 1. Describe all your methods first. Then describe all your results. IMO interleaving your methods and results makes things confusing.

For example, readers seeking to understand your work may not care about your results, and only care about your methods.

1

When writing a scientific paper, be it an undergraduate dissertation or a peer reviewed journal article, the most important thing to remember is your reader is busy and doesn't want to waste time trying to figure out what you're talking about. The answer to this is to heavily signpost everything you do so there can be no ambiguity.

For your particular situation, my approach would be to cover the entire methodology upfront in a methodology section and then when I reach the results section, I would refer the reader to equations 2.1-2.4 for example, for a full understanding of what you did. As long as you are holding your reader's hand the whole way then they'll take the time to actually read and understand what you're doing.

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.