Answers are grounded in this paper and cite their sources.
For the correlation of the boiling points of smaller alkanes we outline the use of the variable connectivity index 1χf involving a parameter x, to be determined during the regression fitting. The optimal values of the variable x is determined by minimizing the standard error of the regression. With the value x=0.65 we obtained a quadratic regression characterized by a standard error of 2°C, when out of 21 structures one structure is eliminated as an outlier (having a standard error over 6°C). This result, based on a single descriptor, is better than hitherto reported regressions on the same set of compounds using other descriptors. Using as an additional descriptor the product P3W3, where P3 are the paths of length three and W3 are the walks of length three, the standard error was further reduced to 1.56°C. We include also the regression results derived for orthogonalized descriptors and clarify the alternative approaches to orthogonalization of descriptors.
No takes yet. Share an insight, caveat, or question.
Milan Randić (2000) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: