# Thread: Effects of general participation and intensity of participation

1. ## Effects of general participation and intensity of participation

Exploring data from a quasi-experimental study, we want to investigate effects of general participation in a program (dichotomous; part_dich) aswell as intensity of participation (continuous; part_cont) on students math-skills (math_slope). More specifically, we wonder if effects of general participation are moderated by intensity of participation. The main problem is, that most students (3/4th) have missings on part_cont (because they did not participate). So far, we came up with two strategies:

1) Only include participating sudents in the analysis and regress math_slope on part_cont (plus several control variables). However, while we can explore the effect of participation intensity within participants, this does not tell us, whether an effect of general participation is moderated by intensity of participation.

2) Split students into groups with low and high participation intensity and calculate separate regression-models (math slope on part_dich plus control variables), both times including non-participants as control group. However, because of the split, we lose more detailled information. Additionally, we seem to have a selection effect regarding intensity of participation, which is not controlled for within separate analyses.

Is there another (maybe even simpler) way to investigate our research-question, which we have not thought of yet?

Lukas

2. ## Re: Effects of general participation and intensity of participation

So you have a variable intensity with values from 0 to maximum. Seemingly you want
to know whether "b2" in the equation y = b0 + b1*intensity + b2*intensity˛ + e is significant?

With kind regards

K.

3. ## Re: Effects of general participation and intensity of participation

Thanks for engaging!

Wouldn't that test a quadratic effect of participation intensity? If I'd compare different interventions, I would rather formulate it like this:

y = b0 + b1*intervention + b2*intensity + b3*intervention*intensity + e

Then, I'd be interested, if b3 is significant, and therefore, if there is an interaction between different interventions and their intensity.

However, since I only have one intervention and a control-group with missings on intensity, this doesn't really seem legit.

4. ## Re: Effects of general participation and intensity of participation

Using the "intervention yes/no" variable is superfluous anyway.
You have an participation ("intensity") variable with minimum 0 and maximum...whatever.
You can analyse whether participation has a linear effect. Or, in addition, whether more
participation has an effect beyond the mere linear effect (therefore, the quadratic term).

With kind regards

K.

5. ## The Following User Says Thank You to Karabiner For This Useful Post:

l.frei (02-10-2016)

6. ## Re: Effects of general participation and intensity of participation

Alright, that would simplify our analysis quite a bit, thanks! I'm still slightly sceptical, because intensity has such a huge floor effect with all those zeros. Could this cause problems or is this mainly a concern for dependent variables?

7. ## Re: Effects of general participation and intensity of participation

Personally, I would start building the model and then check the assumptions and also look for pecularities.
Graphical displays will be useful. Admittedly, I am not sure which particular problems can arise with a
predictor which has 75% zero-values.

With kind regards

K.

 Tweet