Ngene crashes when I run a particular code

[ecd921]

Hello,

I have run the attached code in Ngene requesting 12 rows and it works fine. However, as soon as I request 6 rows (which I could still do), then the program crushes and I ger the following message:

"Something went unexpectedly wrong. You may wish to email Choice Metrics (...) for assistance. It might be something as simple as an error in the syntax that Ngene did not detect, or something a little more tricky. (...) Aborted"

I look forward to hear from you.

Thank you.

Here is the code:

Design
; alts = lambleg, lambchops, goat, beef, none
; rows = 6
; eff = (mnl, d)
; cond:
if (lambleg.origin=1, lambleg.igp=0),
if (lambchops.origin=1, lambchops.igp=0),
if (goat.origin=1, goat.igp=0),
if (beef.origin=1, beef.igp=0)
;con
; model:
U(lambleg) = b1 +
b2[-0.0001] * llprice[4.84, 6.46, 8.07] +
b3.dummy*slaugh[1,0] +
b4.dummy[-0.0001|0.0001]*origin[1,2,0] +
b5.dummy[0.0001]*igp[1,0] +
b6*igp.dummy[1]*origin+
b7.dummy*feed[1,0] +
b8.dummy*lowcarb[1,0] +
b9.dummy*org[1,0] +
b10.dummy*fat[1,0] +
b11.dummy*prot[1,0] +
b12.dummy*pronti[1,0] +
b13.dummy[0.0001]*rouge[1,0]/
U(lambchops) = b14 +
b15[-0.0001] * lcprice[9.32,7.45,5.59] +
b3.dummy*slaugh +
b4.dummy*origin+
b5.dummy*igp+
b6*igp.dummy[1]*origin+
b7.dummy*feed+
b8.dummy*lowcarb+
b9.dummy*org+
b10.dummy*fat+
b11.dummy*prot+
b12.dummy*pronti+
b13.dummy*rouge/
U(goat) = b16 +
b17[-0.0001] * lgprice[10.07,8.06,6.04] +
b3.dummy*slaugh+
b4.dummy*origin+
b5.dummy*igp+
b6*igp.dummy[1]*origin.dummy[0]+
b7.dummy*feed+
b8.dummy*lowcarb+
b9.dummy*org+
b10.dummy*fat+
b11.dummy*prot+
b12.dummy*pronti+
b13.dummy*rouge/
U(beef) = b18[-0.0001] * lbprice[12.06,9.64,7.23] +
b3.dummy*slaugh +
b4.dummy*origin+
b5.dummy*igp+
b6*igp.dummy[1]*origin.dummy[0]+
b7.dummy*feed+
b8.dummy*lowcarb+
b9.dummy*org+
b10.dummy*fat+
b11.dummy*prot+
b12.dummy*pronti+
b13.dummy*rouge/
U(none) = b19
$

[Michiel Bliemer]

We will look into the issue, clearly Ngene should not crash and we will need to update the software, thank you for pointing this out.

I cannot quickly see why this would not work, but it may have to do with the fact that there is simply not enough information 6 rows since many of your attributes are dummy variables (0,1) and therefore across 5 alternatives there are very little trade-offs (many alternatives will have overlapping attribute levels), so while 6 rows satisfies the degrees of freedom in theory, it may not be enough for estimating 20 parameters in your model.

I think that you will need much more variation in your data, e.g. at least 12 or 24 rows would be recommended. You can always block your design.

Michiel