Pearson vs. Spearman – some statistic basics

So I finally got to decide which test to use for my statistical analysis. Unluckily the lecture I took about statistics was in 1999… So I got to research on certain issues now again. But the web is quite a good place to get answers for certain questions, so I see that Pearson vs. Spearman is a quite common issue, So it didn’t take me long to finde some sources to solve my problem:

„The Pearson and Spearman's coefficients are mathematically identical, EXCEPT that the Spearman rank coefficient is calculated from the ranks of each variable, not the actual values. Once the values are ranked, they both use the same formula.” Source.

„Bei intervallskalierten und normalverteilten Proben wird die Produkt-Moment-Korrelation nach Pearson, bei ordinalskalierten und nicht normalverteilten Proben die Rangkorrelation nach Spearman verwendet, welche die Ränge der untersuchten Daten miteinander vergleicht (Bühl & Zöfel 1995, Sachs 1993). Letzteres Verfahren gilt zudem als robuster gegenüber Ausreißern (Sachs 1993). Precht und Kraft (1992) sowie Bühl und Zöfel (1995) weisen darauf hin, dass eine Korrelationsanalyse nur dann sinnvoll ist, wenn der Zusammenhang linear ist; für U-förmige Zusammenhänge tendiert r gegen 0.“ Source.

So it should be better to use Spearman for me, because a) I don’t have to justify why my data should be normally distributed and b) some of the values are not really metrically scaled.

Defined tags for this entry: ,


    No Trackbacks


Display comments as (Linear | Threaded)

    No comments

Add Comment

Enclosing asterisks marks text as bold (*word*), underscore are made via _word_.
HTML-Tags will be converted to Entities.
Standard emoticons like :-) and ;-) are converted to images.
Gravatar, Favatar, Pavatar author images supported.