janvv

janvv

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Background\n",
    "In one of the previous posts, we looked at the maximum likelihod estimate (MLE) for a linear regression model. Instead of using the deterministic model directly, we have also looked at the predictive distribution. In the previous post, we used this stoachstic model to include information about the data uncertainty into the prediction process.  \n",
    "When using the MLE paramter (or the least square estimate in the case of a linear model), all other parameters with smaller likelihood are automatically being discarded. This is problematic because the maximum likelihood parameter, must not be necessarily be correct or even be close to the true parameter value. It simply has the highest data likelihood for the given data. If we have little data points, supporting our hypothesis, the likelihood function becomes very flat. Then, other parameters show very similar likelihood values and the true parameter mig

janvv

janvv