Having found 9 variables of interest plus 3 more variables whose effect should not affect the results of previous ones, can control variables assume either continuous o discrete (dummy) occurrence?
If so, in the final output, should the term "controlling for X control variable" be used?
Thanks in advance.
Hi Nicola,
I am not sure I understand your question? Covariates can certainly be either continuous or categorical.
Additionally, all covariates are considered "adjusted for" in your model, once they are included, so that you could say something like "there was statistically significant effect of X, controlling for the other covariates in the model" (or "when all other covariates are held constant").
I hope this helps
Ariel
If a variable has no statistically significant impact on your outcome variable, you simply say so. You 'control for' variables only when they moderate your model (which results from their correlation with both dependent and independent variables).
However, the separation between predictive and control variables, is not defined by mathematics or statistics: it is driven by your domain theory. Thus, depending on the research hypothesis you are testing, the same variable in the same analysis, can be called predictive, control, confounding or whatever, depending on the spin you are trying to give to your findings.
the variables you control (for example demographics in social sciences) help you to overrule the possibility that for example help seeking behaviour depends on annual income (demographic) rather than levels of PTSD. Both PTSD and income are independent variables and help seeking is dependent variable. What I provided here is strictly "off the sleeve" example to explain the role of predictors.
Hope it helps a bit
- Badges
- Science topic
- Similar topics
- Books