Have an idea?

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

Attribute significance - Logit?

I have a question regarding the interpretation of the logit and latent class analysis.

I would like to say which of my attributes are significant, and if so, p<.05 or p<.01.

Among the output are t-ratios for each attribute level. Can I say something about the significance of the whole attribute, based on these t-ratios per  attribute level?

For example:
1) If two attribute levels are significant t>1.96 and the others are not, can I then automatically say that the whole attribute is significant, with p<.05?
2) If two attribute levels are significant t>2.58 and the others are not, can I then automatically say that the whole attribute is significant, with p<.01?
asked Aug 9, 2017 by Inge

1 Answer

0 votes
To test the entire attribute I would run the model with and without the attribute and use the log likelihood test for significance.
answered Aug 9, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (60,075 points)
...