Preliminary findings on the impacts of the MAUP on t-test: evidence from Monte-Carlo simulation

Keywords: modifiable areal unit problem, hypothesis testing, t-test, size, power
Abstract Type: Virtual Paper Abstract

Abstract

Hypothesis testing is the backbone of inference statistics, based on which solid assertions can be made with due statistical power. One of the most frequently adopted hypothesis tests is t-test for the regression coefficient in a linear regression model, which is used to verify the causality of an independent variable to the dependent variable. However, when spatially aggregated data are adopted for linear regression, the validity and efficiency of t-test are put into question due to the nuisance of the modifiable areal unit problem (MAUP). Monte Carlo simulations indicate that when the null hypothesis is true, the size (probability of making a Type I error) of t-test remains roughly unaffected, with the zoning problem making it more liberal, and the scale problem making it slightly more conservative. However, when the null hypothesis is false, the scale problem significantly undermines the power (one minus probability of making a Type II error) of t-test, which calls for modification of the current t-test to restore its efficacy.