I hoping to examine the effects of sleep on the workplace (e.g. work engagement) using a between-subjects design. I'm having trouble deciding on what to control for.

Let's say there are studies that have found an effect on off-job activities (such as engaging in social activities with friends at night) on work engagement the next day.

There are also other studies on work engagement in particular which controls for job demands and job control as these variables could affect the level of work engagement when it is being measured.

Should researchers control for everything here (off-job activities, job demands, job control)? I'm looking at some papers examining sleep and the workplace and not every possible variable that could affect work engagement is controlled for. Instead others are used such as gender and trait affect.

If there is mediator to the relationship between sleep and work engagement, should the possible variables that influence the mediator be controlled for as well?

I'm guessing I'm really lost about the concept of "control". Can someone please advise? Thank you!