This paper presents the application of runs test for indirect consideration of observation's autocorrelation in estimation of a standard uncertainty of arithmetic mean value. At first stage researches were performed by Monte Carlo (MC) simulation for two kind's random signals: first order autoregression (AR) and moving averaging (MA). Comparison of theoretical values of effective number of observations and determined by the runs test showed a good their convergence (Fig. 2, Fig. 3). This convergence increased when for AR signal level of autocorrelation decreased and for MA signal number of observation increased. To verify simulation results the runs test was used to analyses of random signal of known exponential autocorrelation function (exp (−i/fs·τf)) on output of first order RC low pass filter of time constant τf and different sample frequencies fs. Obtained results (Fig. 8) confirmed simulation results.