A likelihood multiplies the probabilities or densities assigned to observed data under a model. Its logarithm, the log-likelihood, turns that product into a sum. Because the natural logarithm is strictly increasing, the parameter values that maximize the likelihood also maximize the log-likelihood. section 3.1.4
For a normal model with known variance, maximizing this expression leads to the sample mean as the maximum-likelihood estimate. section 3.1.4
The transformation makes optimization easier to express: products become sums, and terms constant with respect to the parameter can be set aside. In logistic regression, maximizing the log-likelihood requires iterative numerical optimization rather than a closed-form solution. section 3.1.4 section 3.3.2
Log-likelihood also connects probability models to familiar fitting methods. With normally distributed regression errors and fixed variance, maximizing it yields the same coefficients as ordinary least squares, which minimizes squared residuals. section 3.4.2



