Dear ACBC User,

We recently discovered a bug in the utility constraints area of ACBC, which actually revealed a "feature" (had we actually documented it, which we didn't...so it hasn't been a feature you've known about yet).

In the documentation, we describe a utility constraints grid wherein relative preferences can be specified for levels (either globally, or customized based on CiW survey responses). We describe that values of 3, 2, 1 assigned to three levels means that the following constraints should be imposed during utility estimation: level 1 > level 2 > level 3. And, we also described that the magnitude of these values didn't matter: 30, 20, 10 leads to the same result as 3, 2, 1.

However, we have discovered a bug: a zero (0) used as a constraint actually leads to no constraint applied for that level! This is the same as leaving that cell in the grid blank.

So, if you have ever typed a 0 (or referred to survey responses with 0 in the data) in the utility constraint grid, then that level was actually not constrained with respect to the other levels within that attribute.

The good news is that if you want to partially constrain an attribute (e.g. constrain level 1>2, but allow levels 3 and 4 to be estimated without constraints), you now know how it is done! You should just leave the cells blank within the constraints grid, for levels 3 and 4.

We will be fixing the bug involving the "0" equating to a "blank" in SSI Web v8.0.2. We will also be fixing this bug in SSI Web v7.0.28. These updates should happen within the next week or two. After the bug fix, the "blank" will still mean to impose no utility constraints for that level (and we'll be updating the documentation to explain this feature.)

We do not have immediate plans to fix this bug in SSI Web v6.

Best regards,

Sawtooth Software