Have an idea?

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

covariates in hb estimation with maxdiff

I have often used categorical covariates in CBC/HB estimation and found them really useful.
I have not done that previously in MAxdiff. Now that maxdiff sets the part-worth of the last item to 0 I am quite confused how to proceed. I received some advice from Sawtooth support but i obviously did not catch how to modify the cho file.
I am sure the info how to do this in maxdiff would be of interest also to others and so pls could you (Brian Orme) repeat the instructions and show how to modify the cho file in the example below. Thank you so much!

4 0 11 18 0
4 1
0  0  0  0  0  0  0  0  0  0  0
0  0  1  0  0  0  0  0  0  0  0
0  0  0  1  0  0  0  0  0  0  0
0  0  0  0  0  1  0  0  0  0  0
2 99
4 1
0  0  0  0  0  0  0  0  0  0  0
0  0 -1  0  0  0  0  0  0  0  0
0  0  0 -1  0  0  0  0  0  0  0
0  0  0  0  0 -1  0  0  0  0  0
4 99
4 1
1  0  0  0  0  0  0  0  0  0  0
0  0  0  0  0  0  0  0  1  0  0
0  0  0  0  0  0  0  0  0  1  0
0  1  0  0  0  0  0  0  0  0  0
4 99
4 1
-1  0  0  0  0  0  0  0  0  0  0
0  0  0  0  0  0  0  0 -1  0  0
0  0  0  0  0  0  0  0  0 -1  0
0 -1  0  0  0  0  0  0  0  0  0
1 99
asked Aug 23, 2018 by merjah (180 points)

1 Answer

0 votes
Any best task (i.e. those that use 1s), need all rows that are all 0s turned into -1s.

Any worst task (i.e. those that use -1s), need all rows that are all 0s turned into 1s.

Implementation of that conversion is potentially tricky, it's possible to do it in Excel by using the "Text to columns" tool and some clever logic to identify and replace rows, or if you know some programming, you could do it with code.
answered Aug 27, 2018 by Kenneth Fairchild Bronze (3,920 points)
...