Hello all,
I conducted a discrete choice experiment and I have two problems:
1) The experiment holds 5 attributes and the ASC. 4 attributes are continuous. The attribute investment concept consists of two levels (own investment and collective investment) which are dummiecoded.
Since the Out Out is "no investment" both of the dummies are coded with 0 in this case. Unfortunetly, this leads to correlation with alterntive specific constant. I tried effect coding but this has not led to any improvement.
Does any one have an idea?
2) I only have a data set of 68 participants in the survey. I am aware, that this is very low and this could be a problem for the estimation. Unfortunately, the sample size can no longer be increased.
Currently I estimate conditional and mixed logit Models. I wondered if there might be another model which allows to better analyze the data.
Many thanks in advance,
Daniel
I conducted a discrete choice experiment and I have two problems:
1) The experiment holds 5 attributes and the ASC. 4 attributes are continuous. The attribute investment concept consists of two levels (own investment and collective investment) which are dummiecoded.
Since the Out Out is "no investment" both of the dummies are coded with 0 in this case. Unfortunetly, this leads to correlation with alterntive specific constant. I tried effect coding but this has not led to any improvement.
Does any one have an idea?
2) I only have a data set of 68 participants in the survey. I am aware, that this is very low and this could be a problem for the estimation. Unfortunately, the sample size can no longer be increased.
Currently I estimate conditional and mixed logit Models. I wondered if there might be another model which allows to better analyze the data.
Many thanks in advance,
Daniel
