Although I don't use R, but I have a few suggestions as follows;

1. I don't know what type of opt-out alternative you have used. For example, there are different types of opt-outs, including 'none of these' or 'no choice' with no levels, that's with zero utility (empty alternative), and 'status-quo' with some current levels, that's with non-zero utility.

2. Whatever may be your situation is, you can create dummy variables, Alternative Specific Constants (ASC) as Cordula Hinkes mentioned to capture the overall effects of alternatives on the observable component of utility other than the effects of attributes of each alternative.

3. Let's say you have 3 alternatives in a choice set, whereas 3rd is 'opt-out', 'none-of-these', 'no choice' or 'status-quo'. Clearly, 'opt-out' will serve as a base alternative. Include this 'no choice' in the number of choices, for example, in the NLogit

choices=3

U (1,2)

This means the number of choices are equal to 3 alternative, however, in the utility function we just included 1st and 2nd alternatives, but didn't include 3rd 'opt-out' alternative so that it can be used as a base alternative.

I hope my answer helps you. If you have more questions, please don't hesitate to ask.

Kind regards

Ok, let's say, I have the following choice experiment with the following choice set design;

* Number of alternatives (ALT) = 4 (4th one is an opt-out or no-choice)

* Number of attributes & levels = 2 Water quality and Biodiversity attributes with 2 levels and 1 monetary attribute with contineous levels as follows;

- Water Quality (WQ): low = 0, high = 1 (0 = reference level),

- Biodiversity (BD): low = 0, high =1 (0 = reference level),

- Entrance fee (EFEE): $ 20, $ 50, $ 70, $ 100, $ 120 (no reference level)

1. Using the following NLogit/Limdep syntaxes, create Water Quality and Biodiversity variables as well as dummy-variables for the Alternative-Specific Constants (ASCs), i.e. for all 4 alternatives, whereas 4th opt-out alternative will serve as a reference or base category:

2. create variables such as low and high water quality variables as follows;

create; if(WQ=0) low-wq; if(WQ=1)high-wq$

2. create variables such as low and high biodiversity quality variables as follows;

create; if(BD=0) low-bd; if(BD=1)high-bd$

3. now, create dummy-variables for all ASCs as follows;

create; if(ALT=1)ASC1 =1$

create; if(ALT=2)ASC2 =1$

create; if(ALT=3)ASC3 =1$

create; if(ALT=4)ASC4 =1$

4. Use the following syntax for MNL (McFadden 's Conditional Logit) model;

sample; all$

clogit;lhs=choice; choices=1,2,3,4;

Model: U(1,2,3)

=

+high-wq*high-wq

+high-bd*high-bd

+efee*efee$

In choices = 1, 2, 3, 4, whereas in the utility, we have alternatives 1, 2,3 because alternative 4, which is an opt-out or no-choice will serve as a base category.

In variables, we used only high-wq and high-bd variables because low-wq and low-bd will serve as base categories. However, for entrance fee a reference or base is not required since it is a countinous variable. For example, for gender (male = 0, female =1) and employment (not employed =0, employed=1) reference level is zero, whereas for continuous variables like income ($ 1000, $ 2000, $ 3000, etc.) and age (18, 25, 30, 35, 40, etc.) we don't need a reference level. So, the same is the case of our monetary attribute or entrance fee.

I hope this will greatly help.

Kind regards