By Anonymous on Monday, July 21, 2003 - 02:04 pm:
Hi,
what is the acceptable limit for a non zero intercept?
In a similar topic someone said that the intercept should be within the 95% confidence limit of the curve, in my case, my placebo is ok (no peak) and my intercept is 10% of my greater standard concentration, the correlation is fine R^2= 0,996 but when I use the 0,0 as a data point it falls into 0,96.
If I donīt have a matrix contribution in by peak why I donīt have a zero intercept?
Is there some technical answer?
By Daren on Wednesday, July 23, 2003 - 09:00 am:
Hi,
I have always followed the guidelines that y=0 has to be withing the 95% confidence bands of your curve in order to force through zero. You should have at least 5 std. points on your curve when you run this or else you will be artificially broadening your confidence intervals. I would even stack more points down at the low end of the curve to make sure you get an accurate picture of your true intercept. As long as a zero intercept falls within your 95% confidence lines you can have a single point curve that is forced through zero, if zero is outside of your confidence lines than you will have to use the mult. point curve when running this analysis.