quadratic growth curve question

[ExitG]

Hello,   In most nutrition studies, the classical nutrient response curve is not quadratic.  And that is what my research group is encountering now.  I have 2 sets of growth response data from flowering plants.  The quadratic model is barely significant at the 10% level and the correlation is poor.  The optimum and confidence intervals it predicts are over-estimates.  Remember, I want the lowest fertilizer concentration that generates an optimum growth response.  Just by looking at the data, I would pick that level to be about 0.25 to 0.5 mM in both experiments.  Any model that does not go this low, does not fit the biology. So my question is, can we model such a curve.  On the vinca data, a 6th order polynomial generates a pretty good best fit curve.   Can one do with higher order polynomial, or some other line model, and do the same confidence interval determination as you showed us how to do with quadratic? I have the data in Excel and would send it to anyone who can help me with this question.   Thank you,   Carrie