So you have a file that contains all of these independent variables and you know they are not all important for your model. Which ones do you drop? Which ones do you keep? Hmmmm. Well, I would drop variable X at a glance because I am not concerned with it. Hold up…The way statistics works is this…you want to test each variable for significance and confounding. You want to know the relationship between the variables you collected because someone found them to be noteworthy. So my suggestion is to test the model for all variables (known to be relevant) and eliminate from that point.
What you want to avoid is excluding the wrong variables. You don’t want variables in the model that serve no purpose either….So the ideal situation is you capture a model that only has relevant significant factors – and then in your write-up you explain quantitatively how they relate to the outcome variable.
If you enjoyed this short blog there are others to follow.-Amy
