what is a non-linear regression?
I am just asking because econometrics/Mathematical economics is perhaps the hardest thing to understand, and I don't even know where to start.
To know what a
non-linear regression is we need to know what a
linear regression is first.
The purpose of running a linear regression is to find values for a slope and intercept (in high school algebra it was presented as y = mx+b) that can help one fit a line through your observational data.
More formally in Econometrics it is specified as:
Yi = B0 + B1Xi + Ui
i = [1,n]
Yi is the independent variable.
Xi is the dependent variable.
B0 + B1Xi is the population regression function.
B0 is the intercept of the line.
B1 is the slope of the line.
Ui is the error term.
More precisely, the linear regression program finds values for the slope and intercept that define the line that minimizes the sum of the square of the vertical distances between the points and the line.
An example of Econometrics using a linear regression is Okun's law which shows the relationship between the GDP growth and change in unemployment is approximately linear. This data set shows the law in application to US quarterly data.
However many relationships in economics and other feilds (like biology for example) do not follow a straight line. The relationship is either logarithmic, quadratic, hyperbolic... etc. This is where non-linear regression comes into the picture.
Nonlinear regression is a general technique to fit a curve through your data. It fits data to any equation that defines Y as a function of X and one or more parameters just as above. It finds the values of those parameters that generate the curve that comes closest to the data (minimizes the sum of the squares of the vertical distances between data points and curve). Again the same as above.
The generalized model is Yi = X'iB + Ui. But can take forms like ln(y) = ln(a) + bx.
i is a (row) vector of predictors for the ith of n observations, usually with a 1 in the first position
representing the regression constant.
B is the vector of regression parameters to be estimated.
Ui is a random error.