1
$\begingroup$

Does anybody know how to interpret a whole bunch of effects (main and interaction) in a clever way? Or does anybody have a good example where it's shown?

To be more precisely: Assume that you have a lot of effects in your model (main and interaction effects) and you know that standard errrors and coefficients are biased (as they are in most empiric studies). But you want to present results, what would you do? So because of the bias p-values are wrong, so there is no valid way to say which coeficients are significant and which are not. So there has to be another way to present information. The idea to work with plots seems to be a good one. But I'm still curious how you guys handle such problems!

$\endgroup$
2
  • 2
    $\begingroup$ I'd suggest looking at graphical summaries of the estimated effects. Here are two papers that give suggestions, (Gelman et al, 2002; Kastellec and Leoni, 2007) $\endgroup$
    – Andy W
    Jun 15, 2011 at 17:39
  • 1
    $\begingroup$ you'd like elaborating your question a bit more. $\endgroup$ Jun 16, 2011 at 5:40

1 Answer 1

1
$\begingroup$

I like the Predict.Plot and TkPredict functions in the TeachingDemos package for R, but my opinion may be slightly biased.

Here is an example (the TkPredict function allows you to dynamically change values to see how they compare): enter image description here

$\endgroup$
3
  • $\begingroup$ Hello Gregg! I don't really understand what's going on in this plot. Thanks so far! $\endgroup$
    – MarkDollar
    Jun 16, 2011 at 7:42
  • $\begingroup$ That is just one of the example plots. The 3 lines in this case are the 3 different species of irises and the lines are showing the effect of Petal.Width for a constant value of teh other variables. You can use the functions mentioned to create the multiple lines at different values of the covariates to show how the predictions change. With interactions you can see the shape change as well as shifting. The TkPredict function lets you interactively change the values at which you are predicting to see the change. $\endgroup$
    – Greg Snow
    Jun 16, 2011 at 17:27
  • $\begingroup$ Hmm this function is just giving me a Scatterplot?! $\endgroup$
    – MarkDollar
    Jun 18, 2011 at 7:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.