---
title: "Çoklu Regresyon Analizi Varsayımları"
url: "https://merterbil.com/papers/coklu-regresyon-analizi-varsayimlari/"
author: "Mert Erbil"
pdf: "https://merterbil.com/papers/coklu-regresyon-analizi-varsayimlari.pdf"
---

# Çoklu Regresyon Analizi Varsayımları

English title: Assumptions of Multiple Regression Analysis

A methods paper by Mert Erbil, from his Ph.D. studies, Advertising & Public Relations (dropped out), Bahçeşehir University, Istanbul. Written in Turkish.

## Abstract

Multiple regression is one of the most common analyses in social sciences, and it's easy to run it in SPSS without asking whether your data allows it. So I wrote this methods paper during my Ph.D. studies to put those assumptions in one place, in Turkish, with the exact steps to test each of them.

It starts with what regression actually means (relating one variable to one or more others) and what a residual is, the part the model can't explain. Then it splits the assumptions into two groups.

The first group is checked before the analysis, on the data itself. So it covers sample size (a rule of thumb of 50 plus 8 per independent variable), normal distribution with visual and statistical tests, linearity through scatterplots and outliers.

The second group is checked after the analysis, on the residuals. That means independence of residuals with the Durbin-Watson test, constant variance (homoscedasticity), normality of residuals with a P-P plot, multicollinearity with tolerance and VIF values, and outliers that mislead the model, using Mahalanobis and Cook's distances.

The last part is a sort of cheat sheet. It lists the SPSS menus and options for every check, plus a sample data set. And with it you can go from raw data to a model you can defend without guessing which box to tick.

Full text (PDF): https://merterbil.com/papers/coklu-regresyon-analizi-varsayimlari.pdf

All papers: https://merterbil.com/papers/