Dear Ngene team,
I am a PhD student. For my second study I am doing choice experiment. I have some confusion. I did pilot test with 50 participants and did conditional logit model to get the coefficients. But still I am not clear if I needed to do another analysis. My question is that ok to do conditional analysis to get the coefficients? And what is the reason to do conditional logit analysis with pilot data?
Kind regards
Shama
Question about pilot
Moderators: Andrew Collins, Michiel Bliemer, johnr
-
Michiel Bliemer
- Posts: 2055
- Joined: Tue Mar 31, 2009 4:13 pm
Re: Question about pilot
The purpose of a pilot study is two-fold:
1) Ensuring that you will be able to estimate models using the pilot data
2) Obtaining informative priors for generating a (Bayesian) D-efficient design; this is optional
To establish both, you estimate a multinomial/conditional logit model using the pilot data, assess whether all coefficients have the correct sign and make sufficient sense. Then you can use the parameter estimates and standard errors to determine Bayesian priors to generate a Bayesian-efficient design that you can use in the main study to make the data collection more efficient. But if you are happy with the experimental design in the pilot study then you can keep it and you do not need to generate a new design.
Michiel
1) Ensuring that you will be able to estimate models using the pilot data
2) Obtaining informative priors for generating a (Bayesian) D-efficient design; this is optional
To establish both, you estimate a multinomial/conditional logit model using the pilot data, assess whether all coefficients have the correct sign and make sufficient sense. Then you can use the parameter estimates and standard errors to determine Bayesian priors to generate a Bayesian-efficient design that you can use in the main study to make the data collection more efficient. But if you are happy with the experimental design in the pilot study then you can keep it and you do not need to generate a new design.
Michiel