What's the standard terminology for variables in observational studies vs experiments?

Question

My stats textbook (De Veaux et al., 2008) advises using the terms response variable and predictor/explanatory variable in observational studies vs response variable and factor in experiments. However, I've also found comments on Quora that suggest that the term response variable is inappropriate for observational studies because it implies a causal relationship. In my field, sociolinguistics, in which most studies are observational, I've also been advised to use the terms dependant variable and independent variable, though I've also read arguments against the term independent variable in observational studies because it implies that the researcher has the ability to establish a control and manipulate the independent variable.

My field itself is moving towards using response variable and factor, which seems like an odd mix, unless I've been misled about what the common terminology for variables in observational studies is. So, since I'm attempting to stick to standard terminology for ease of reading by non-sociolinguists and since I want to avoid any confusion about whether my current study is able to establish causal relationships or not, which terms should I use?

(As far as the sort of variables I'm using, I have one in particular which I'm analyzing against all the others in order to discover statistical associations. e.g. I want to know if age or sex or location, etc., are associated with one's choice between synonyms.)

Answer 1

Unless you are trying to write a textbook you might be better off avoiding such terminology issues. For your study, the "response variable" (or whatever you would want to call it in general) of interest is evidently "synonym choice." Use that specific phrase throughout when you write up your results and then you don't have to worry whether the word "response" is appropriate. The word "factor" has a particular standard meaning for describing a categorical as opposed to a continuous variable, so that probably shouldn't be used unless all of your "predictors" are categorical. The word "feature" is reasonably unaffected by these terminological conundrums and nicely covers categorical, ordinal, and continuous "predictors."

Answer 2

The terminology differs between fields, unfortunately. So I suggest that you

1. stick to the convention within your field, and
2. define what you mean by the term in your methods.

Incidentally, there is one more option for 'dependent/response variable', and that is target variable. I find predictor and target (which may be standard in machine learning) to be simple and clear while avoiding any implications of causal relationships.

Answer 3

In the analysis code I write I call them "observational variables," which I think is a good term. But I don't see it being use anywhere else.