When I use Lasso in a linear regression, how do I use the coefficients to predict the dependent score? E.g., if the dependent score ranges continuously form 0 to 100 and I have 4 non-zero coefficients in the model, but I have no constant, how do I use these coefficients to predict the absolute score? I'm not interested in saying that Person A will score X points higher than Person B, because A has a factor that B doesn't.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
The same way as for ordinary least squares: if your estimates of the parameters are $b_1$, $b_2$, … $b_n$, then the model's best guess of the dependent variable for a person with covariate values $x_1$, $x_2$, …, $x_n$ is $b_1x_1 + b_2x_2 + \cdots + b_nx_n$. (If you had an intercept $b_0$, which I presume is what you mean by "constant", you'd just add that on.)
answered Aug 11, 2017 at 17:43