Effect of modulating braille dot height on reading regressions