Date:         Sun, 25 Sep 2005 13:03:01 +0200
Reply-To:     Marta García-Granero
              <biostatistics@terra.es>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Marta García-Granero
              <biostatistics@terra.es>
Organization: Asesoría Bioestadística
Subject:      Re: MANCOVA data transformation
In-Reply-To:  <20050923.231812.548.4.babyblue325@juno.com>
Content-Type: text/plain; charset=ISO-8859-15

Hi Jordan,

JH> I am analyzing the data for my dissertation and I've got a Box's M on one
JH> of my DV's (I can tell which one, due to outliers on each end). Since the
JH> DV is a ratio, arcsine has been suggested to me as a transformation.

Are you talking about a percentage or just the ratio of two
quantitative variables?

Arcsin transformation is suitable for percentages (or fractions, but not
any ratio) that are likely to follow a binomial law (instead of the
normal law), provided that the denominator is a constant. If lack of
HOV is caused simply by some outliers, but not because your data are
binomially distributed, then it is  not a good idea at all to arcsin
transform them.

JH> do you think would be best? And can I only transform one of the two DV's,
JH> or do I have to do both? Also, once I transform the data, what other
JH> things do I need to keep in mind when looking at other parts of the data?
JH> Meaning, the means of the DV etc. These have to be converted back, yes?

Some transformations can be undone, that's why I consider a bit
dangerous to use then without a solid reason.

Regards,
 Marta                            mailto:biostatistics@terra.es