Chapter 9 Simple Regression
🚧 Under Construction 🚧
We assume you are familiar with regression or Ordinary Least Square (OLS), but let’s make a quick recap on what regression is all about.
9.1 Univariate regression
is a regression between one dependent variable and one independent variable. Most textbooks use a notation \(Y\) for dependent variable, and use \(X\) as the independent variable. A univariate regression assume this form: \[ Y_i=\alpha+\beta X_i+ e_i \] where \(Y_i\) is your observed dependent variable and \(X_i\) is your observed independent variable.
There are many versions of the Gauss-Markov assumption, but really there are two main thing we assume for OLS to be unbiased:
- error term \(e_i\) have a conditional zero mean, i.e., \(E(e_i|X_i)=0\)
- error term \(e_i\) is independent and identically distributed
We call \(\alpha\) as intercept. We generally don’t interpret \(\alpha\) although it is very important to include \(\alpha\) to avoid bias. We are interested in \(\beta\), as it shows the general strength of relationship between our \(X\) and our \(Y\) is. That is, we often say:
an increase of \(X\) by 1 unit, is associated with an increase of \(Y\) by \(\beta\) unit, assuming everything else constant.
9.1.1 running your univariate regression
suppose you have
9.1.2 plotting your regression
In the previous chapter, we have learned how to plot our data using
plot(). You can add to your script a line showing your regression result
9.2 Multivariate regression
Multivariate regression is just like your univariate regression, but we have more independent variable than just one. Dependent variable still only one. That is:
\[ Y_i=\alpha+\beta X_i+\gamma Z_i+ e_i \] The interpretation is similar.
9.2.1 running multivariate regression
Now, instead of
unfortunately, you can’t plot multivariate regression as this regression has more than two dimension.