Have an idea?

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

How is it possible that not all main effects are significant (count), but all attributes are (|t|>1.96 - logit)?

How is it possible that the count analysis indicates that some main effects are not significant, but the t-ratios of the logit analysis indicate that all attributes are significant?

The sample size is +/- 20
20% of the main effects do not seem to be significant in the count analysis
All attributes are found to be significant in the logit analysis.
asked May 6, 2020 by Mike

1 Answer

0 votes

The counts analysis is an extremely preliminary thing you might want to look at.  It is  not a model and it doesn't take into account the experimental design.  Disconnects between it an any statistical model of the data are numerous and for any time the counts report says one thing and a statistical model says another, you're better off to trust the statistical model.
answered May 6, 2020 by Keith Chrzan Platinum Sawtooth Software, Inc. (105,750 points)