Regressions/Models

Once you graph/analyze your data, how can you tell if it will be an exponential model, linear model, or power model? Help please!

0) a priori expectation. It is not all that unliekly that the researcher should have a clue before starting.
1) Experience provides an excellent initial guess.
2) Fit the data! Which provides the best fit?
3) Different measures of "fit" may produce different results.