choice-metrics.com

Hi there,

I've run an Ngene syntax where I optimise on an MNL model with 6 Bayesian priors and evaluate for an RP panel model with 1 random parameter with 2 Bayesian priors and 4 other Bayesian priors.
I've let the syntax run for half a day (about 12 hours) on Ngene version 1.2.1 and use the last evaluation in the iteration history.

When I double click the last evaluation in the iteration history, it brings up the design for that evaluation and I can see the properties for the MNL efficiency measures and the design matrix.

However, when I expand the tab: Design properties (rpppanel_model) and click on the RP Panel box to see the RP Panel efficiency measures, Ngene crashes...I think?
The Ngene screen greys out and the top left of the Ngene project window has the following text next to the Ngene icon: Ngene (Not Responding).

Why is this happening please?
How can I fix this please?
I've spent 2 half days running syntax and obtaining an output at the end of each design run, but I can't access the RP panel efficiency measures for each of these designs. I'm running out of time to get the design output for my main study.

Any help with this would be gratefully appreciated please.

Thanks,
Rakhee.

Without knowing the exact syntax it is difficult to tell what is going on. Note that when clicking on rppanel it may be doing millions of calculations, which may take quite some time to finish (if it actually fits in memory) and Ngene will show "not responding" while it is actually busy calculating.

For example, if you are doing 100 Halton for Bayesian priors (set via ;bddraws), 100 draws for random parameters (set via ;rdraws) and a simulated sample of 500 (set via ;rep), Ngene needs to conduct 100*100*500 = 5,000,000 simulated covariance matrices, which may take a long time.

You could try setting ;bdraws = gauss(3), ;rdraws = gauss(3), and ;rep = 200.

Michiel

Hi Michiel,

Thank you for responding.

I've included my syntax below:
Design
;alts (mnl_model) = GP, RNP, PP
;alts (rppanel_model) = GP, RNP, PP

;rows = 20
;block = 4
;eff = mnl_model(mnl,d,mean)
;bdraws = gauss(3)
;alg = mfederov
;require: GP.X1 = 1

;model (mnl_model):
U(GP) = b1.effects[-0.023] * X1[0,1] + b2.effects[0.176|0.017|0.003] * X2[0,1,2,3] + b3[(n, -0.244,0.026)] * X3[0,1,2,3] + b4.effects[-0.083|-0.029] * X4[0,1,2] + b5.effects[0.030] * X5[0,1] + b6.effects[-0.024|0.049] * X6[0,1,2] /
U(RNP) = b7[(n, -0.230,0.066)] + b1 * X1 + b2 * X2 + b3 * X3 + b4 * X4 + b5 * X5 + b6 * X6 /
U(PP) = b8[(n, -0.347,0.075)] + b1 * X1 + b2 * X2 + b3 * X3 + b4 * X4 + b5 * X5 + b6 * X6

;model (rppanel_model):
U(GP) = b1.effects[-0.006] * X1[0,1] + b2.effects[0.217|-0.003|0.008] * X2[0,1,2,3] + b3[n,(n, -0.304,0.046),(u,0.208,0.420)] * X3[0,1,2,3] + b4.effects[-0.084|-0.033] * X4[0,1,2] + b5.effects[0.020] * X5[0,1] + b6.effects[-0.013|0.036] * X6[0,1,2] /
U(RNP) = b7[(n, -0.291,0.075)] + b1 * X1 + b2 * X2 + b3 * X3 + b4 * X4 + b5 * X5 + b6 * X6 /
U(PP) = b8[(n, -0.401,0.081)] + b1 * X1 + b2 * X2 + b3 * X3 + b4 * X4 + b5 * X5 + b6 * X6
$


You indicated that by clicking on rppanel it may be doing millions of calculations, which may take quite some time to finish (if it actually fits in memory).
Question 1: How do I know if it actually fits in the memory please?

When I expand the Design properties (rppanel_model) tab and click on MNL, or RP or OOD box, I can see the efficiency measures for these models.
The problem happens only when I click on the RP Panel box within the Design properties (rppanel_model) tab.
Question 2: Should I just click once on the RP Panel box within the Design properties (rppanel_model) tab and then leave the desktop alone and let it run the calculations?
Question 3: Will it eventually display the RP panel efficiency measures when it is done calculating?
Question 4: How long would be too long to wait and know that there is a problem?


Many thanks,
Rakhee.

1. Ngene/Windows will crash
2. Yes you need to be patient, it may take a long time.
3. Yes unless it runs out of memory
4. Sometimes it is a few seconds, sometimes a few minutes, sometimes hours, it depends on how many draws you are doing, this is your choice. If you add ;rdraws = gauss(5) and ;rep = 200 then it should finish calculations within a few minutes. The more draws the more accurate but the higher the calculation times.

Michiel

Thank you for the explanations Michiel.
I'll give it more time to do the calculations for the RP Panel evaluation.
If Ngene/Windows crashes or it's taking too long, as suggested I'll use ;rdraws = gauss(5) and ;rep = 200 instead of the ;bdraws = gauss(3) that I have in my current syntax.

Many thanks,
Rakhee.

i did not suggest to remove ;bdraws, please keep it, I suggest adding the two commands I mentioned.
You are currently using a lot of Halton draws for random draws (the default) but gauss(5) is more efficient. If you want to use more draws, increase the gauss abscissas and specifically increase ;rep to eg 500 or 1000.