1: \begin{abstract}

2: We have tested the leading correction--to--scaling exponent $\omega$

3: in \linebreak $O(n)$--symmetric models on a

4: three--dimensional

5: lattice by analysing the recent Monte Carlo (MC) data. We have found that

6: the effective critical exponent, estimated at finite sizes of the

7: system $L$ and $L/2$, decreases remarkably within the range of the

8: simulated $L$ values. This shows the incorrectness of some claims that

9: $\omega$ has a very accurate value $0.845(10)$ at $n=1$.

10: A selfconsistent infinite volume extrapolation yields

11: row estimates $\omega \approx 0.547$, $\omega \approx 0.573$, and

12: $\omega \approx 0.625$ at $n=1$, $2$, and $3$, respectively, in approximate

13: agreement with the corresponding exact values $1/2$, $5/9$, and $3/5$

14: predicted by our recently developed GFD (grouping of Feynman diagrams) theory.

15: We have fitted the MC data for the susceptibility of 3D Ising model at

16: criticality showing that the effective critical exponent $\eta$ tends

17: to increase well above the usually accepted values around $0.036$.

18: We have fitted the data within $[L;8L]$, including several terms in the

19: asymptotic expansion with fixed exponents, to obtain the effective

20: amplitudes depending on $L$. This method clearly demonstrates that

21: the critical exponents of GFD theory are correct

22: (the amplitudes converge to certain asymptotic values at $L \to \infty$),

23: whereas those of the perturbative renormalization group (RG) theory are

24: incorrect (the amplitudes diverge). A modification of the

25: standard Ising model by introducing suitable "improved" action (Hamiltonian)

26: does not solve the problem in favour of the perturbative RG theory.

27: \end{abstract}

28: