Different samples give different results

Even when a sampling method is sound, separate random samples often produce different estimates. One sample may contain 42% walkers and another 47%. This sampling variation is expected. Larger well-designed samples generally give more precise estimates, but size alone cannot repair systematic selection bias.