Hi,

I've recently been re-acquainting myself with linear algebra and can admit that econometrics isn't my best subject. I'm trying to prove the existence of a "Kuznet's Curve" relationship between my regressors and regressand. The model uses panel data (cross sectional, time components). An earlier study models the relationship as:

(1) y_{it}= b1 + b2.x_{it}+ b3.x_{it}^{2}+ b4.z_{it}+ u_{it}

where x = IV, with k1 x 1 matrix,

z = IV, with k2 x 1 matrix

u(t) = error term, normally distributed, mean 0, variance of o^2.

My questions are:

1. Does the above model represent a bi-variate quadratic equation?

- various sites list the general form of a bi-variate quadratic as (2)

- My assumption is that equation 1 is obviously a bi-variate quadratic, much in the same way that y=x^{2}is still a quadratic equation (general form of y=x + x^{2}+c)

2. Are the illustrations of bi-variate quadratics correct in the below link?

Curvilinear Regression

Or more importantly, the b4 coefficient of equation 1 has a linear relationship with y, and will affect the steepness of the curve.

I think the answers are pretty obvious (yes it is a bi-variate quadratic), but lack of sleep has made me illogical made me-over think things! Would appreciate the help!!