I have a linear function $y=\alpha + \beta k + \beta x$ and observation data that consist of pairs of $x$ and $y$. $\alpha$, $\beta$, and $k$ are unknown parameters. I want to estimate the value of $\alpha + \beta k$ and $\beta$. I don't need the values of $\alpha$ and $k$. Is it possible to solve this problem by using the Ordinary Least Squares estimator? Or if there are easier ways to solve this problem?
Asked
Active
Viewed 993 times
1 Answers
0
Is k another variable or is it constant?. If it is a constant, lump the first two terms into , one big variable called K and use ordinary least squares method to estimate K and beta with your labeled data (as you do not care about alpha and k). If it is not, then you need to know the value of k's in order to use OLS to figure out the coefficients. and keep the model $y=\alpha + \beta(k+x)$ and your new pairwise labeled data would be $(x_i+k_i,y_i)$
Satish Ramanathan
- 11,984
-
$k$ is an unknown constant parameter. So it is ok to use OLS as you said even when $K$ and $\beta$ are not independent? – Yan Li Jan 02 '16 at 05:13
-
I believe you are OK. but I would not say that K and $\beta$ are independent, even in a regular regression, K and $\beta$ are dependent. You are good to go. – Satish Ramanathan Jan 02 '16 at 05:30