I would be grateful for any advice on the following statistical
analysis. I am trying to compare the slopes and elevations of three
regression lines (for three species, A, B and C). However, X is not very
accurate (an average) which contravenes one of the assumptions of
regression analysis, that of the X values being measured without error.
During preliminary analysis I performed linear and polynomial
regression on the data. To determine the " best fit" regression line, I
used a stepwise procedure, increasing the polynomial degree by 1 and then
calculating an F value to see whether the increase in degree of polynomial
significantly improved the accuracy of the fit. I found that for species
A, a 2-degree polynomial was sufficient and for species B and C, a linear
regression line produced a good enough fit. However, on examining the
residuals Vs fitted values, the values did not appear to be randomly
dispersed. This can be seen from a simple scatter diagram of the data.
There appears to be a less variability in the data at the higher values of
X.
My questions are:
Could anyone advice me on whether there is a more appropriate model for the
data I
have ?
How do I compare between straight and curved lines to see whether there are
any significant differences in regression line elevation and/or slope?
Many thanks in advice for your help,
Cassie James
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news