2-in-1 design

[peyman_07]

My decision is to set all priors to zero. Previously, you mentioned that even if I use no priors, ASC0 must have a value. Should I use those insignificant priors for ASC0? For instance Bayesian priors, for M1: ASC0[(n,-0.259,1.24)] and for M2: ASC0[(n,2.18,2.36)].

[Michiel Bliemer]

That is not what I wrote. I quote: "You can never leave ASCs as zero unless all other priors are also zero."

[peyman_07]

Hi,

I just created a model averaging design but needed to change some attribute levels manually. Then, I wanted to evaluate the design with the following syntax in Ngene:

design

Code: Select all

;alts(m1) = ACC*, DEC*
;alts(m2) = ACC*, DEC*


;rows=24
;block=6
;eff = m1(mnl,d) + m2(mnl,d)

;eval=H:\My Documents\eval.csv


;model(m2):


U(ACC)= tr.dummy[0]*TR[1,2]+rt.dummy[0|0]*RT[1,2,3]+wt[0.0001]*WT[0,5,10,15]+dr.dummy[0.0001]*DR[1,2]+rr[0.0001]*RR[3,4,5]+pt[-0.0001]*PT[5,10,15,20]+sp[0.0001]*SP[0,1.5,3](10-12,8,4-6)+pl.dummy[0]*PL[1,2]+tcg[-0.0001]*TCG[0,15,30]+gt[0.0001]*GT[0,1.5,3]+tf[0.0001]*TF[10,15,20]/

U(DEC)= ASC0


;model(m1):

U(ACC)= tr.dummy[0]*TR[1,2]+rt.dummy[0|0]*RT[1,2,3]+wt[0.0001]*WT[0,5,10,15]+dr.dummy[0.0001]*DR[1,2]+rr[0.0001]*RR[3,4,5]+pt[-0.0001]*PT[5,10,15,20]+sp[0.0001]*SP[0,1.5,3](10-12,8,4-6)+lo.dummy[0.0001]*LO[1,2](7-9,15-17)+pl.dummy[0]*PL[1,2]/

U(DEC)= ASC0

$



When I run this syntax, this error appears: "ERROR: More designs found in data that in model specification."

Would you please let me know what it means?

Kind regards,
Peyman.

[Michiel Bliemer]

It likely means that there is an error in the format of your CSV file.

Please consult the Ngene manual for the appropriate way to format a design, namely:

* there needs to be a header column
* the first column indicates the design number, typically a column with ones only, i.e., 1, 1, 1, etc.
* the second column indicates the choice task number, i.e., 1, 2, 3, etc.
* the remaining columns list the attribute levels in the order in which they appear in the utility functions

Michiel