Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

How to conclude an attribute is significant in the logit or latent class analysis?

I am identifying whether attributes are significant in explaining the choice for the logit and latent-class analysis. When the t-ratio of one of the attribute-levels is significant, does this mean this attribute is significant in explaining choice? And how can I determine the significance - level of the attribute afterwards?
asked Jun 27, 2017 by SanderM01 (160 points)
edited Jun 27, 2017 by Walter Williams

1 Answer

0 votes

Good question.  The t-ratios tell you which attribute LEVELS are significant.  If you want to know if an entire ATTRIBUTE is a significant predictor of choice or not I think the easiest way would be to run the model twice, once including the attribute and once excluding it.  Then compare the log likelihood of the two models using the log likelihood test.  This gives you a direct measure of the significance of the attribute as a whole.
answered Jul 2, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (66,325 points)