I am doing a regression analysis. I have taken several variables as per the literature (say for example 15 classified as firm character, financial, governance) . I have also taken one variable which is not found in literature, but for which i have an intuition (say, variable,X). Now when I am trying to fit the model with all the 16 variables, i have to make iterative process to arrive at one model where there is a perfect fit. (say , for example , with 6 variables). Now out of these 6 variables, 4 variables have significant p value (including the variable X which I added out of intuition). Now I have the following doubt:
In my hypothesis development, should I include all the 15 variables which I originally considered for the model? or should I take only the six variables which eventually gave me a model fit?