Accurracy of 10-point Newton-Cotes formula

Hi everyone,

my problem is not diectly related to MKL software and I think that this foum is more appropriate for asking question about the numerical integration.I 'am developing a numerical integration library which is based on well known multipoint Newton-Cotes formulas taken directly from Mathworld article.I have implemented already most of the formulas and ran simple accurracy related tests which gave a satisfactional results.In  case I have encountered a problem which is probably related to catastrophic cancelleation because of alterating values(points) beign used to achieve the convergence of Integral.Can anyone help me?I would like to know how to minimize the error which is always ~0.2.

Thanks in advance.