The multinomial Logit model (MNL) is suitable if the set of alternatives is (i) complete (containing all possible options), exclusive (only one option can be chosen), and (iii) the alternatives are sufficiently different from each other such that correlations can be neglected. For example, "bus" and "train" are both public transport with many commonalities (fixed time table, fixed stations etc) and therefore correlated. For such cases, a "nested logit model" is suitable. Notice that the binomial logit model is just a special case of the MNL for 2 alternatives.
We are almost neighbours - I live on the Sunshine Coast. I assume you are asking what is a statistical tool for the analysis of various options to get to the airport (ie. Airtrain, self drive car, friend drop off, taxi, bus, etc).
I would expect that factors impacting the decision could include time of day, business versus pleasure travel, distance from the airport, wealth of the person, number of people in the travel group, etc.
When comparing a range in mixed categorical and numerical independent variables with a number of categorical dependent variables, I believe Factor Analysis is a valid approach - at least initially.
I am keen to make contact with the UQ Transportation and Infrastructure researchers so perhaps we could get together at some stage and discuss this and other topics with your supervisors and colleagues.
I think it is based on no. of mode choice available for the users.. If it is two modes than Binary Logit Model can use and if it is more than two modes than Multinomial logit model can use...
The multinomial Logit model (MNL) is suitable if the set of alternatives is (i) complete (containing all possible options), exclusive (only one option can be chosen), and (iii) the alternatives are sufficiently different from each other such that correlations can be neglected. For example, "bus" and "train" are both public transport with many commonalities (fixed time table, fixed stations etc) and therefore correlated. For such cases, a "nested logit model" is suitable. Notice that the binomial logit model is just a special case of the MNL for 2 alternatives.
I would suggest you to use mixed logit models which allows the investigation of mode choice including correlated options. This is possible by introducing random parameters.
I do recommend you to read the book written by Train on discrete choice modelling.
You can estimate these model using biogeme a free software you can download from the Internet .