Is there a way to increase power in my study

[Rick_1]

Hi everyone,

I'm working on my 'behavioral change' master thesis. My goal is to design an intervention to reduce printing at the workplace in a big organisation. Unfortunaltely i dont get provided with the printing details on a personal level, so my best bet is too see how much a printer printed in a certain period.

I want do a pre and post-intervention measurement from two weeks each. Also i want to work with two control floors and two intervention floors. So I can control for confounds. But here's my problem: there are only 20 printers in the organisation. So my sample is n = 20. Is there a possibilty to increase power by increasing the number of measurements per printer (maybe on a daily basis?)? Or something else to increase my power?

With kind regards,

Rick

 

[hlsmith]

Are you able to randomize the intervention? Ideally that would eliminate differences between areas, but you would still want to examine for baseline differences between printer usages in area to make sure these covariates aren't confounding results given the small sample. You would also want sufficient number of values to ensure represented observation patterns related to your data generating process (people printing). You may be able to do a interrupted time series with a control group. That design would all for multiple measures and a control group.