Hello?
Does anyone know how I can implement the below equation in R? I would
like to estimate the following equation:
y=beta_ij * (1+gamma_j * dummy) * x_ij
where y is continuous, and all the x variables (j of them) are i=3
level categorical variables. The intuition is that instead of
estimating the additive value for a dummy variable, I would like to
estimate the multiplicative value for the dummy variable. Thus the
presence of the dummy would scale the beta. Note that for each x
variable there is only one gamma.
For concreteness, you can imagine that y is a continious test score, x
are categorical variables indicating different types of education
achievements, each type of education achievement is categorised in 3
levels (none, some, a lot), and the dummy indicates race. In this
model I believe that race affects test scores proportionally to
estimated beta of each education level. This avoids having to estimate
a gamma for each education achievement level.
Is the solution to simply use nls {stats} and type out the equation?
Hope the explanation makes sense, happy to explain further.
Best wishes,
Peter
How to setup a multiplicative dummy function in R
2 messages · lolo koko, Jeff Newmiller
2 days later
I think you need an offset term, or maybe I just don't understand your question. A sample data set, particularly if you can show us how your equation could be used to generate the sample data, would be helpful.
Sent from my phone. Please excuse my brevity.
On December 1, 2016 7:22:37 AM PST, lolo koko <lokomiauw at gmail.com> wrote:
>Hello?
>
>Does anyone know how I can implement the below equation in R? I would
>like to estimate the following equation:
>
> y=beta_ij * (1+gamma_j * dummy) * x_ij
>
>where y is continuous, and all the x variables (j of them) are i=3
>level categorical variables. The intuition is that instead of
>estimating the additive value for a dummy variable, I would like to
>estimate the multiplicative value for the dummy variable. Thus the
>presence of the dummy would scale the beta. Note that for each x
>variable there is only one gamma.
>
>For concreteness, you can imagine that y is a continious test score, x
>are categorical variables indicating different types of education
>achievements, each type of education achievement is categorised in 3
>levels (none, some, a lot), and the dummy indicates race. In this
>model I believe that race affects test scores proportionally to
>estimated beta of each education level. This avoids having to estimate
>a gamma for each education achievement level.
>
>Is the solution to simply use nls {stats} and type out the equation?
>
>Hope the explanation makes sense, happy to explain further.
>
>Best wishes,
>
>Peter
>
>______________________________________________
>R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>https://stat.ethz.ch/mailman/listinfo/r-help
>PLEASE do read the posting guide
>http://www.R-project.org/posting-guide.html
>and provide commented, minimal, self-contained, reproducible code.