0
$\begingroup$

I am conducting a literature review and noticed that some studies on my topic (drivers of telecommuting) only analyze their survey data using correlation analysis instead of regression. I am wondering why someone might decide just to analyze survey data using correlation analysis when they can take the next step to analyze the data using regression analyses?

$\endgroup$

closed as off-topic by mkt, Michael Chernick, Peter Flom Feb 9 at 11:30

  • This question does not appear to be about statistics within the scope defined in the help center.
If this question can be reworded to fit the rules in the help center, please edit the question.

  • 2
    $\begingroup$ Either they are only interested in the bivariate (marginal) association of two variables, so not in the effect of a single coefficient controlled for all other covariates on an outcome, or they do not know about regression analysis. $\endgroup$ – tomka Feb 8 at 16:03
  • $\begingroup$ TY! I guess I still don't understand under what circumstances would they only decide just to publish results showing the association when they can move forward with testing the effect on the outcome variable? Does it have to do with sample size? $\endgroup$ – user3424836 Feb 8 at 16:19
  • 1
    $\begingroup$ Have a look at @whuber 's answer here: stats.stackexchange.com/a/294341/32477 Maybe that will help figuring out why the authors did what they did given the context of those studies you mention. Since people here don't know the context, you won't likely get a satisfactory answer to your question. $\endgroup$ – Stefan Feb 8 at 16:31
  • 3
    $\begingroup$ One advantage of the correlation analysis is that you don't have to designate one variable as dependent and one as independent. With surveys this might be an appropriate approach. If people who live further from work telecommute more, which is the independent variable and which is the dependent? $\endgroup$ – Sal Mangiafico Feb 8 at 16:46
  • 2
    $\begingroup$ Also, if you have several (potentially treated as independent) variables that are likely to be correlated, using them as independent variables in a multiple regression analysis will mask the relationship between some of these and the dependent variable. For example if you had both a) distance to work, and b) time to commute to work, as your independent variables, a multiple regression analysis might select only one of these to be significant or have a large effect. Better to just present all the correlations among all variables, and let the reader decide? $\endgroup$ – Sal Mangiafico Feb 8 at 16:56