Regressions conducted with transformed data

Appendix 7 Regressions conducted with transformed data... 

There are many cases in which a researcher would like to map out the entire regression model (including both bo and hi) with a 1-a Cl. If the data have excess variability, the Cl will be wide. In fact, it may be too wide to be useful. If this occurs, the experimenter may want to rethink the entire experiment or conduct it in a more controlled manner. Perhaps more observations—particularly replicate observations—will be needed. In addition, if the error (y — y) = e values are not patternless, then the experimenter might transform the data to better fit the regression model to the data. 

The methods for data treatment included first-derivative transformation of log (1 /T) data with window size of 25 points, based on the Savitzki-Golay (22) polynomial filter. Calibration for quantitative determination of log SCC and absolute electrical conductivity was performed with PLS regression as described above. 

---