TABLE 3.

Complete Data: Students in classes that use warm RC are more likely to report increased participation by themselves (columns 1 & 2) and their peers (columns 3 & 4).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
	I Participate		Others Participate		Motivated		Community		Comfort Asking
Warm RC(ref: non-RC)	0.660 (0.255)0.0097	0.545(0.167)1.11e–03	1.860(0.266)2.85e–12	1.710(0.175)1.43e–22	null	null	null	null	null	null

The estimates are presented as log-odds, standard error of estimate in parentheses, p-value below. (Note that backwards model selection was performed using AICc to identify the best fitting model so the p-value should be interpreted with caution, if at all.)

All models only include an indicator for warm RC and a random effect (even columns).

Odd columns show effects when including a random intercept for student and unique instructor in a unique class.

We show both estimates with and without the random intercept to demonstrate that they do not vary considerably; for simplicity of interpretation, we plot the effects in probabilities from the models that do not use random effects (even column numbers).

When the null model is the preferred model, “null” replaces estimates.

Note that a null model is a model that only includes an intercept (no predictors) and random effects (if applicable).