Hi everyone,
I have finished the data collection of my ACBC (10 attributes, 26 levels, summed price approach).
For HB, I am not a 100% sure whether I should include interactions. I read the manual and I am aware that it should only be done if really necessary.
If I run the interaction search tool with a linear price function. I get 57 values, 36 are significant with a p-value below 0,005. For some, the ‘Gain in Percentage Certainty over Main Effects’ is super low. As I have read in several forum posts, I can ignore them. But I have 6 interactions that indicate a ‘Gain in Percentage Certainty over Main Effects’ of over 10%, the highest is 22%. How should I treat them?
Run
+ Ela x Water
Parameters in Model
22
Log-Likelihood Fit
-27.041,42
Chi Square Value
133,54851
2LL P-Value for Interaction Effect
0,00000
Gain in Pct. Cert. over Main Effects
22,05%
However, since I want to use a piecewise price function, I also tried a different approach. I first ran HB analysis with piecewise price function. Then I used ‘Export Utilities’ to export an HBU file and used this file for another interaction search. I get 57 values, 50 are significant with a p-value below 0,005. In addition, I get some super high values for the ‘Gain in Percentage Certainty over Main Effects’. The highest is actually 78%. Is this possible or did I do something wrong?
Run
+Dur x Wei
Parameters in Model
5
Log-Likelihood Fit
-17.199,86
Chi Square Value
475,53691
2LL P-Value for Interaction Effect
0,00000
Gain in Pct. Cert. over Main Effects
78,53%
Also, strangely, the ‘Charts’ tab says: No charts are available since there are no significant interactions for both runs.
Thanks in advance for your advice.
Happy holidays
Michael
