Explanatory variables that cannot be controlled or fixed: Does the regression work?
Linear regression analysis is one of the most used statistical techniques in experiments planned to study the functioning of natural systems, especially in measurable studies. Many times, the researcher does not have the ability to control the explanatory portion of the regression model, so the explanatory variables can be as random or more than the response variable. This could generate biases in the estimates of the associated slopes and lead to wrong conclusions. An alternative to the classical regression method is type II regression when the values of the explanatory variable cannot be controlled. This paper presents different situations based on published research in ecology and agronomy for different purposes: prediction, estimate of the slope and comparison of slopes between two groups, in which the problem of random variation in the explanatory variables is present with different degrees of relevance. In each case, the most appropriate path for the analysis will be identified. A simulation was also carried out that considered different combinations for the random errors in the regressor and response variables in order to visualize the bias of the estimators in each situation for the different regression methods. It is clear from the foregoing that it is necessary to emphasize two very important issues in order to decide the most appropriate type II regression method: be clear about the objective of the work and if the application conditions required by each method are met. This review aims to be a simple guide to when and what method to apply in each situation.
| Main Authors: | , , |
|---|---|
| Format: | Digital revista |
| Language: | spa |
| Published: |
Asociación Argentina de Ecología
2020
|
| Online Access: | https://ojs.ecologiaaustral.com.ar/index.php/Ecologia_Austral/article/view/1066 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|