Even although growth-curve fashions allow evaluation of how results develop, they don't seem to be proof against confounding from unmeasured variables that may have an effect on both the preliminary.