In my research in a plant propagaton study, I found root length
significantly affected by the treatments tested.  When analyzing the
data of treatments A, B, and C, I found treatment B to be the best
treatment in terms of root length (means separated by LSD at alpha <
0.01).  However, when producing a standard curve using regression
analysis it appears visually that treatment C (or possibly higher) is
the best treatment in terms of root length. Should the regression
curve always be highest at the treatment B if that is the best
treatment found by analyzing the data from the experiments?  What
should I look for to explain this?

---

Request for Question Clarification by jeremymiles-ga on 21 Feb 2003 07:49 PST

Did you dummy code the data for the regression?

jeremymiles-ga

What stat tests are you using to analyze the data?  Have you plotted
the "error about the curve" for the regression line?  How good is the
"fit" for each regression line?