How analyze annual data with one IV and multiple DVs

Question

I'm currently working with a data set that includes multiple variables associated with each of 10 years of data. The basic structure, with (example hypothetical) variables in caps, is from YEAR to year, TV WATCHING increased, which led to less HAPPINESS and less SLEEP.

I want to test whether TV watching increased as time went on, and whether those increases led to a decrease happiness and sleep. I have found a statistically significant positive correlation between YEAR and TV WATCHING, and a significant negative correlation between TV WATCHING and HAPPINESS/SLEEP. But I would like to do a more sophisticated analysis.

Please let me know what you feel would make the most sense. I'm thinking I could do a multivariate regression with TV WATCHING as the single IV and HAPPINESS/SLEEP as the two DVs. But I'm not sure how robust that would be, plus it ignores the time element. So then I was thinking structural equation modeling with YEAR -> TV WATCHING -> HAPPINESS; SLEEP. But I wasn't sure that was the best approach either.

Any assistance will be much appreciated. Thanks!

Answer 1

Your description sounds like a hypothesis of mediation, if you want to search that literature. Path analysis (structural equation modeling) is ideally suited to estimate all the paths of interest simultaneously. For example, using the R package lavaan, the model syntax might look like:

model <- ' WATCHING ~ YEAR
           HAPPINESS + SLEEP ~ WATCHING '

You can label parameters to estimate indirect effects of YEAR on each outcome via TV watching, as illustrated in the tutorial example: https://lavaan.ugent.be/tutorial/mediation.html

Of course, if you measured the same participants across years, then the nested structure of the data must be accommodated (repeated measures nested within subjects). There are many ways to approach that, if necessary (e.g., multilevel SEM or latent growth models).