Talk:Introduction to Chemical Engineering Processes/Basic Statistics and Data Analysis

Linearization in order to do linear regression is a poor (and antiquated) approach to data interpretation. Relative errors are distorted in ways that can give very bad fits to the equation in the pre-linearized form. The most extreme example is when the inverse is taken, as is often recommended with "hyperbolic" forms such as Michaelis-Menten. Students should be advised, at least, to plot the model equations vs. data in their original form and axes, to check for this. Better still is to learn non-linear regression, which is simple to do on any spreadsheet with Solver capability.