A scheme of regression applying

Introduction. If You'd like to clear up how some variable (response, Y) depends on a set of other variables (regressors, predictors, X), You may use regression methods. Such methods differ one from another in many ways. You should consider Your data's nature (measure type and distribution type, first of all). Below, a scheme of regression applying is presented which helps to choose the fitting regression method.

##### Is the response's measure type scale?

##### For identifying a variables' measure type, learn this topic (in Rus.). After identifying, click the suiting button on the right.

##### Clarify whether the regressors as main effects explain the response enough

##### For clarifying it, apply an analysis of variance as a criterion for regressors' strength when using them as main effects (the series "Analysis of variance and linear regression with categorical regressors. Primary exploring a functional link's shape" containing 9 videos). After clarifying, click the suiting button on the right.

##### It is useless to apply any regression method to the exploring regressors as main effects because the expected model will produce a forecast which accuracy will be lower than a modal forecast's accuracy. Then, You may try interaction effects or apply just correlation analysis or try another system of regressors and start from the beginning.

##### Including interaction effects in a regression model may heighten its accuracy. According to a thumb rule, it may rise to roughly a half of the initial accuracy (i.d., the accuracy of the regression model with regressors as main effects). Hence, if Your model's initial accuracy (R-squared) equals 0.01, You hardly may gain the accuracy (R-squared) of 0.50 after including interaction effects in the regression model.

##### Clarify whether the regressors as main effects and interaction effects explain the response enough

##### For clarifying it, apply an analysis of variance as a criterion for regressors' strength when using them as main effects and interaction effects (the series "Loglinear analysis. Primary exploring a deep interaction structure" containing 10 videos). Interaction effects

#####

##### After clarifying, click the suiting button on the right.

##### It is useless to apply any regression method to the exploring regressors as main effects and interaction effects because the expected model will produce a forecast which accuracy will be lower than a modal forecast's accuracy. Then, You may apply just correlation analysis or try another system of regressors and start from the beginning.

##### Dichotomize the categorical regressors containing more than two categories

##### You may apply full or partial dichotomizing. When full dichotomizing, any category of a regressor becomes a new binary variable. When partial dichotomizing, some category of a regressor are joining. Choosing between full and partial dichotomizing, You may apply common sense, theoretical frame, and analysis of variance as a criterion for variables' categories similarity (the series "Analysis of variance. Implementing the found out deep interaction structure" containing 8 videos). After dichotomizing, click the button on the right.

##### Identifying a functional link's shape

##### For identifying a functional link's shape, apply an analysis of variance to as a criterion for functional link's shape (the series "Analysis of variance and linear regression with categorical regressors. Primary exploring a functional link's shape" containing 9 videos). After identifying, click the suiting button on the right.

##### Implement linear regression modeling and exam its results

##### For identifying a functional link's shape, apply an analysis of variance as a criterion for functional link's shape (the series "Analysis of variance and linear regression with categorical regressors. Primary exploring a functional link's shape" containing 9 videos). After identifying, click the suiting button on the right.