### Betas for orthogonal polynomial

Posted:

**Mon Jun 28, 2010 8:12 am**Hello all,

I have six waves of election studies from which I created a long format dataset with a categorical identifyer "year". I model the effect of the year using orthogonal polynomial transformation (linear and quadratic) and interact predictors in the model with the "year" in order to get specific estimates for each time-point. However, in MLwiN I can only get estimates for the middle of my time-trend. Namely, if I have six waves from 1987 to 2007 I only get the estimates for 1997 - the year when no election was held. Of course, I can do customized predictions for other years, but then I need to have ten graphs rather than one simple table. Essentially, what I need is the beta coefficients and standard errors for all other years and I am not sure if/how to get them.

Note: Predictions window is not an option b/c my independent variables are continious so in this case I am getting an estimate for each value rather than an overall effect.

Thanks,

Dmitriy

I have six waves of election studies from which I created a long format dataset with a categorical identifyer "year". I model the effect of the year using orthogonal polynomial transformation (linear and quadratic) and interact predictors in the model with the "year" in order to get specific estimates for each time-point. However, in MLwiN I can only get estimates for the middle of my time-trend. Namely, if I have six waves from 1987 to 2007 I only get the estimates for 1997 - the year when no election was held. Of course, I can do customized predictions for other years, but then I need to have ten graphs rather than one simple table. Essentially, what I need is the beta coefficients and standard errors for all other years and I am not sure if/how to get them.

Note: Predictions window is not an option b/c my independent variables are continious so in this case I am getting an estimate for each value rather than an overall effect.

Thanks,

Dmitriy