0
$\begingroup$

I'm concerned that I may not be conceptualizing/applying multiple regression correctly; I would greatly appreciate some guidance.

Lets say that I'm trying to understand how to get a TV audience to watch more of my TV show.

In other words, I'm trying to influence my audience's motivation to watch.

A multiple linear regression has shown me that overall enjoyment of the show accounts for ~60% of the variance in the audience's motivation to keep watching. So it seems that if I increase the overall enjoyment, that will hopefully increase motivation to watch.

So how do I increase overall enjoyment? I then computed another multiple regression, which showed that a single good joke in an episode accounts for ~70% of the episode's overall enjoyment rating from the audience.

However, in my original regression, the regressor of 'single good joke' did not account for any of the variance in audience motivation.

So I have a sort of A>>B>>C situation; single good joke >> overall enjoyment >> increase motivation

Is it prudent to suggest that we should focus on producing one good joke per episode in order to increase enjoyment, and therefore increase audience motivation? I'm concerned because 'good joke' does not directly account for 'motivation' variance, so I'm not sure if I'm being ignorant of the logic of regression.

Any tips are appreciated. Thanks!

$\endgroup$
5
  • $\begingroup$ This sounds like a problem that could be solved with Structural Equation Modeling. $\endgroup$
    – Tim
    Dec 2, 2014 at 12:48
  • $\begingroup$ Was "single good joke" an IV (instrumental variable; as you state) or just a simple regressor in your original regression? $\endgroup$ Dec 2, 2014 at 16:53
  • $\begingroup$ @RichardHardy 'single good joke' was just a regressor. $\endgroup$
    – Nick
    Dec 2, 2014 at 20:51
  • $\begingroup$ Could you perhaps make it clear in the question text as well? Because there the IV of 'single good joke' still stands. You could write "regressor" instead of "IV". It might help avoid unnecessary confusion. $\endgroup$ Dec 2, 2014 at 21:48
  • $\begingroup$ Would you care to write down the regressions you used? If you have more than just the three variables that you explicitly mentioned, it could help convey the overall picture better. Perhaps the single_good_joke variable is jointly significant with some other variable, for example (just a guess). $\endgroup$ Dec 2, 2014 at 21:53

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.