<p>In regression studies, especially in general linear models, variable selection is key. It is vital for statistical inferences. Recently, Chatterjee’s correlation has been introduced as a novel measure of dependence. This criterion can measure the nonlinear or functional relationship between two variables. On the other hand, the general linear model does not necessarily require a linear relationship between the response variable and covariates. This paper proposes a new algorithm that selects the appropriate variables by sequential tests for the general linear model based on Chatterjee’s correlation. In classical algorithms, changing the functional form of the general linear model may lead to the selection of different covariates. But, in the new algorithm, selected variables aren’t affected by changing the functional form of the general linear model. Also, a theorem discusses the algorithm’s properties, and simulations show the excellent performances of the new proposed algorithm. Finally, we applied our method to real data.</p>