the more large sample size the more prior not significat

[Tiziana]

Dear all
I did today my second pilot and the resuls are totally opposite with the first pilot. Since I believe that there is high hetergeneity inthe sampl, I pulled the data and estimating a MNL or RPL most of coefficient are not significant. Moreover the D error increased rispect to previous results. What it the correct efficient design? Which of both pilots is right? I am sure that people have responded not in creasy way, but just because most people like a chips with salt other ones without salt or fat.
What do you suggest me, now?
Thanks in advance
Tiziana

[Michiel Bliemer]

There are many things that could lead to insignificant parameter estimates:
(1) your sample size is too small
(2) the respondents do not understand the choice tasks
(3) the respondents do not think that the attributes are very relevant
(4) the attribute level range that you are using are too narrow

Note that studies that investigate low priced goods that people purchase often (like potato chips), people do not really think much about the choice, it is mostly habitual. This in contract to high priced goods that people purchase only once every few years (like cars), people will really trade off attributes.

I do not know your study and I do not know what came out of any literature studies or focus groups you have done to investigate which attributes are relevant and what attribute levels are relevant, I can only comment on your syntax in Ngene, which is fine. I am afraid I cannot answer any questions regarding specific studies.