A related issue is that of model validity. When we discuss the validity of a model, we typically mean if the specified model does what it is intended to do. In other words, does the model actually model what we are looking to model? To answer the question, we look at the internal validity of a model and its external validity.
The internal validity of a model, such as multivariate regression, looks at whether the independent variables used are statistically significant; that is, are the internal constructs of the model valid? We typically use the p-value from regression to measure this internal validity. The null hypothesis tested is that each of the independent variables has zero effect on the dependent variable. Hence, low p-values, at or below the alpha significance level, imply that it is statistically significant and impacts the dependent variable. Hence, a model with only significant independent variables is deemed internally valid.