In time series analyses assessing the relationship between risk factors and respiratory diseases or mortality, influenza incidence is a potential cofounder and, therefore, must be controlled. Available influenza data come from weekly compulsory notifiable disease (EDO system). Furthermore, its graphical distribution, suggests that information may be underestimated in holiday periods. We have applied a procedure for estimate daily influenza series from the weekly cases, using less, a non-parametric local fit. Findings show that smoothing could avoid stepping and sudden peaks in the original series.