 What is bias we have a linear regression where we want to estimate the parameter of theta? Let's say that its true value is theta star now if we take thousands of examples of x and y We can estimate the value of theta hat once now Let's repeat this process a few times where we take different sets of x and y's and have different estimates of theta hat if You take the average of all of those estimations of theta hat Let's say that you get a value that's over here Then this distance over here becomes the bias Ideally we would want the distance between this yellow line and green line to be zero if that were to be the case Then this linear regression up here would be an unbiased estimator for the point estimate theta