Using nonparametric verus parametric tests after transforming variables???

[bubs83]

Hi! I'm a bit confused. I've transformed a subscale due to it being abnormal however because the other 3 subscales in the overall scale are normal (and thus not transferred) do I use parametric tests with the transformed subscale? On the flip side IF i need to use non parametric tests I use the non transformed subscale?

Please help!

Thanks

[hlsmith]

Interesting questions, I may lean toward just using the nonparametric for all three variables with the non-transformed data. I believe that it would be fine, but hopefully those with objections will provide their feedback.

[bubs83]

Thanks for your insight

[Karabiner]

I've transformed a subscale due to it being abnormal

There's no reason for transfoming sample data just because they are
distributed in such a way that seemingly they do not originate from
a normally distributed population. With some statistical procedures,
the distribution of the residuals is of concern, but almost never the
distribution of th original values. Moreover, "normality" concerns
diminish with increasing sample size (unfortunately, sample size was
nowhere mentioned).

With kind regards

K.

[bubs83]

There's no reason for transfoming sample data just because they are
distributed in such a way that seemingly they do not originate from
a normally distributed population. With some statistical procedures,
the distribution of the residuals is of concern, but almost never the
distribution of th original values. Moreover, "normality" concerns
diminish with increasing sample size (unfortunately, sample size was
nowhere mentioned).

With kind regards

K.

Hi there,
Thank you for your input. Our uni's psych statistician advised our group to transform the subscale so we just went along with it. The sample size was 157.