ApolloChoiceModelling forum

Dear Concern,

I have gone through the code examples in the Apollo forum. I wanted to use the mixed logit I order to capture the preference heterogeneity in the sub-population and hence I would like to use MMNL/Mixed logit model. But I have a question regarding the starting value. Citing an example from the code given in Apollo website:

### Vector of parameters, including any that are kept fixed in estimation
apollo_beta = c(mu_log_b_tt = -3,
sigma_log_b_tt = -0.01,
mu_log_b_tc = -3,
sigma_log_b_tc = -0.01,
mu_log_b_hw = -3,
sigma_log_b_hw = -0.01,
mu_log_b_ch = -3,
sigma_log_b_ch = -0.01)

I am pretty much not sure, how these starting values are selected based on the data of Swiss Metro. In other words, What kind of starting values should I use? Is it flexible to choose the stating values?

Hi

this model uses lognormal distributions, and using -3 as the starting value for the mean of the log of the coefficient (which is Normal) implies that the starting value for the coefficient is close to 0. See the FAQ section on the website too

Stephane