ApolloChoiceModelling forum

Tim the sign of the correlation depends on the sign of the product of hinc_sig_20* hinc_sig_20_90. Section Section 9.2.5 in Kenneth Train's book. The Delta method can be used for this, but you'd need to work out the derivatives yourself - they're not implemented in Apollo for the standard errors. Be...

Hi Tim I've looked into your model in detail and I believe I have found the reason this is happening. In your base specification, you use the same coefficient whether the treatment is 20 or 90. In your more general model, you use separate parameters. Now, you see that in the MNL model, the more gene...

Hi Tim thanks for sharing the code and data with me which allowed me to investigate this. On my machine, the code runs fine with Sobol and with other draws. The error you are getting is actually not by Apollo but by randtoolbox which is called by Apollo for Sobol draws. Can I suggest that in the fir...

Hi Richard on point 1, what you do is indeed the easiest and best way of doing it. You could use the same list for the different components, but you would need to reset it in between each component (so you start with an empty one). While that would save a little memory compared to creating new lists...

Hi Houman

thanks, this was a bug. This has been fixed in v.0.1.1 which you can download from the website (http://www.apollochoicemodelling.com/code.html)

Best wishes

Stephane

Tim that's a much larger difference in fit than one would expect with different draws but is in line with recent tests I've done on antithetic draws and changing signs of sigmas. What happens when you use your model with different draws, but use the results from the reduced model as starting values ...

Tim I wasn't necessarily implying that you should use the same draws for increases and decreases, but that this could be a reason for the difference in fit. Using different draws arguably makes more sense, but let's try and understand what you're seeing a little better. 1. What draws are you using, ...

Hi Tim let us assume you have the following situation: model 1: V = beta_x * x + ... model 2: V = beta_xinc * x_inc + beta_xdec * x_dec + ... where x_inc = (x>x0)*(x-x0) and x_dec = (x<x0)*(x-x0) not that in this notation, x_dec will be negative, which will help us with the comparison As only differ...

Tim

could you please also share the output?

Thanks

Hi Tim

Could you please share your code and the full output you see so we can investigate?

Thanks

Stephane