Hello, may I ask how "effect coding" and "estimate coding" are distinguished in scripts? ( U=b1*B1+b2*B2....)
When attributes are all continuous varibles, means that neither "dummy" or "effect" are present.
About :design coding & estimation coding
Moderators: Andrew Collins, Michiel Bliemer, johnr
-
Michiel Bliemer
- Posts: 2062
- Joined: Tue Mar 31, 2009 4:13 pm
Re: About :design coding & estimation coding
Design coding is typically used for orthogonal designs these designs are not optimised for specific values of the utility function, so any levels can be chosen by the analyst and design coding is most common. For example, b1 * colour[0,1,2].
For generating efficient designs, all attributes need to be coded exactly as you would estimate the model. This means for numerical attributes using the actual levels that you would use in your data, e.g. b1 * traveltime[10,20,30] for travel times in minutes. For categorical attributes you MUST you dummy or effect coding, e.g. b1.dummy * colour[1,2,3], which is translated into b1(1) * (colour == 1) + b1(2) * (colour == 2). The values 1,2,3 can be chosen to be any values. What is important is that you add ".dummy" or ".effects" to convert the levels into dummy or effects coding.
Michiel
For generating efficient designs, all attributes need to be coded exactly as you would estimate the model. This means for numerical attributes using the actual levels that you would use in your data, e.g. b1 * traveltime[10,20,30] for travel times in minutes. For categorical attributes you MUST you dummy or effect coding, e.g. b1.dummy * colour[1,2,3], which is translated into b1(1) * (colour == 1) + b1(2) * (colour == 2). The values 1,2,3 can be chosen to be any values. What is important is that you add ".dummy" or ".effects" to convert the levels into dummy or effects coding.
Michiel