Page 1 of 1
Specifying hurdle MXL model
Posted: 07 Dec 2022, 19:54
by Esraa
Hello,
I would like to run MXL model with hurdle component to address serial non-participation for those who "did not play the game".
Can you please advise on how to specify the hurdle framework in Apollo?
Many thanks,
Esraa
Re: Specifying hurdle MXL model
Posted: 12 Dec 2022, 16:58
by stephanehess
Hi
could you be more specific, maybe with some equations?
Thanks
Stephane
Re: Specifying hurdle MXL model
Posted: 13 Dec 2022, 14:15
by Esraa
The idea of the hurdle model is to use random utility to model respondents who show little utility-evaluating behaviour (non-participants) - and that requires a specification of the behaviour of serial nonparticipants (SNPs).
I came a cross a good paper that explains a blended approach that combines the latent class model and the hurdle model, where the hurdle structure addresses the serial non-participants, while the latent class modelling detects clustered patterns within participants’ response set. I would like to do the same thing with a mixed logit model (instead of latent class).
Here is the paper,
https://link.springer.com/article/10.10 ... 20-00434-z
but I will do my best to explain it here as well.
In the paper, the authors specified that a serial non-participant as those who always chooses the status quo. If probability of being
an SNP is modelled with a probit model, we get:
Pr {SNP} = Φ (Zn *se)
where the variance of the residual normalized to 1 and individual characteristics denoted with Zn.
The hurdle framework represent the possibility of a subject always choosing the status quo (SQ) by a summation the two parts:
Pr {All SQ} = Pr {All SQ|P} ∗ (1 − Pr {SNP}) + Pr {All SQ|SNP} ⋅ Pr {SNP}
Re: Specifying hurdle MXL model
Posted: 02 Feb 2023, 13:39
by stephanehess
Hi
sorry about the slow reply
So essentially, this sounds like a two stage model. You could specify both models as components in apollo and multiply them together using apollo_combineModels
Stephane