Coarse-grained models for linear flexible polymers are constructed defining effective segments by taking together n successive chemical monomers of a polymer chain, for n=1,2,3,... . The distribution function Pn(l) for the length l of such effective segments is studied as well as the distribution function Pn(ϑ) of the angle between successive effective segments. If n is large enough, all these distribution functions tend towards universal limiting functions. For small n, information on chemical structure and effective potentials for the various degrees of freedom of the polymer chains is still preserved. Using polyethylene (PE) as one example, it is shown that these distribution functions for small n depend somewhat on the choice of the model for the effective potential (and the degrees of freedom included). Bisphenol-A-polycarbonate (BPA-PC) as a second example, serves to study to which extent these distribution functions Pn(l) and Pn(ϑ) differ for chemically different polymers, such as PE and BPA-PC. Consequences for the molecular modeling of polymeric materials are briefly discussed.