It's interesting to view the discussion on curve fitting. I think it is important for students to understand such "parent" functions as quadratic, exponential, logarithmic, etc.
The AP Course Description speaks of making "transformations to achieve linearity; logarithmic and power transformations." This is an appropriate reason for students to graph the data first to determine its probable parent function. Then the students can transform the data to make to output liner. For example, If the data appears to be quadratic and the output is in LIST2 then LIST3 would be the sqrt(LIST2). The plot then becomes linear and can use a linear regression Sqrt(y) = a+bx. Then y = (a+bx)^2. This preserves the true idea of linear regressions and allows students to see how residuals are an important tool for analyzing the projected model. It is really the same idea as using log-log paper or semi-log paper.
There are several good activities that provide hands on investigation for these type problems. I plan to use them so student understand the mathematics behind the transformations as well as the statistics. Larry Peterson Bonneville High School Home 251 E Laker Way 4470 S 1650 W Ogden, Utah 84405 Roy, Utah 84067 801-476-5808 801-731-4267