uneven attribute levels and number of levels

[socu15_2]

Hello,
I'm designing an unlabeled choice experiment using the m-federov algorithm.
The syntax below has produced a design with perfect attribute balance, each price level present twice per block.
I wanted to ask two questions

1) Is it a problem that the price levels are not evenly spaced? I've heard contradictory advice on this and wanted to make sure it's fine in this case...
The reason for choosing the price levels as below is that there's quite a bit of inequality in the sample, with some consumers being a lot more responsive to price than others, so I wanted to avoid a range that was too small but also introduce some smaller price gaps, if that makes sense.

2) Is it likely to be enough for each price level to appear 4 times in the overall design (twice per block)?

I hope the rest of the design is fine...
Thanks a lot in advance

Design
; alts = alt1*, alt2*, alt3
; rows = 24
; block = 2
; eff = (mnl, d)
; alg = mfederov(candidates = 2000)
; con
; model:

U(alt1) = b1[0.0001] + b2[-0.0001]*price [60, 65, 70, 75, 80, 90](4,4,4,4,4,4)
+ b3.dummy[0.0001]*low_salt[1,0] (10-14, 10-14)
+ b4.dummy[0.0001]*safety_certified[1,0] (10-14, 10-14)
+ b5.dummy[0.0001]*no_antibiotics[1,0] (10-14, 10-14)
+ b6.dummy[0.0001]*originSA[1,0](10-14, 10-14)

/
U(alt2) = b1 +b2*price
+ b3*low_salt
+ b4*safety_certified
+ b5*no_antibiotics
+ b6*originSA

$

[Michiel Bliemer]

1) No, not a problem at all. One usually prefers equally spaced attribute levels such that they cover the entire attribute space more or less equally but for model estimation it does not matter.

2) Yes that is enough, you are not using dummy coding for the price attribute so that means that each level does not need to appear a large number of times, you are only estimating a single price parameter.

Michiel

[socu15_2]

Great, that makes sense, thanks a lot!