Backward elimination stepwise multiple regression model for distance and DSP was developed using variables found to be significant (p < 0.05) in the univariate analysis.
In these cases, a significant difference was noted for DSP, but not distance shown in Table 2 and Table 3. Significant variables identified by univariate analysis were assessed in a stepwise multiple regression model, using backward elimination method.
In an alternative model we included only the last born singleton children (N = 4051) and ran multiple regression model with stepwise backward elimination including the variables: wealth status, development region, sex of child, number of antenatal care, iron supplementation consumption, tobacco smoking, mother's education, mother's occupation and type of cooking fuel (Not shown in table).
Accordingly, we used a backward stepwise regression procedure to identify the most appropriate multiple regression model based on a starting model that included age, TCDD concentration category, and TCDD concentration × age.
Regression coefficients, associated confidence intervals, and P values of start multiple regression model and final model after stepwise backward selection.
To construct the final multiple regression models, backward step-wise linear regression has been used.
Next, multiple regression models were fitted, employing backward elimination of non-significant terms beginning with the maximal model.
The multiple regression models for AEE, sleeping HR, and grip strength were performed using backward elimination including selected predictor variables only.
A backward elimination procedure was then used to discard all predictor variables with P < 0.1 in multiple regression models, one by one, until a final 'best' model was achieved.
We fitted multiple regression models.
