A model is chosen based around its ability to follow the general trend that you see in the data. the correlation coefficient and coefficient of determination can also be used to give some guidance as to what is the best model for the data given. We also need to consider how appropriate it is when predicting outside of the data. Note: some non linear models increase/ decrease at an ever increasing rate which means predictions may not be so good too far away from the data.

Note: all models are limited to how far they should extend by what is a reasonable prediction. You need to explain what values your model would be good to predict over eg some variables can't go negative or even close to 0 like weights of humans etc. Also there will be a limit as to how high a variable can go before it is unreasonable human weights.