When you have no obvious reference category or control group, rather than use dummy coding you can use effects coding which makes comparisons with the grand mean instead. You can either use weighted or unweighted effects coding if you have unequal group sizes, the latter taking into account the sample size of your groups. You should find lots of info on the internet about these coding systems, and various stats papers that address the issue.
Dr. Kathryn Gardner
School of Psychology
University of Central Lancashire
Tel: +44(0)1772 89 4463
[log in to unmask]
>>> Fiona Kennedy <[log in to unmask]> 12/09/08 12:39 PM >>>
I hope someone can help me with what i think should be a straightforward query (i may be wrong thou)
I've got a sample of breast cancer patients who have received different surgical treatments - primarily having a WLE (ie. area removed) or mastectomy, but then some received reconstruction and others did not.
Originally i entered these into SPSS as two variables:
1) Final surgery - WLE or mastectomy (1, 0)
2) Reconstruction - Yes or No (1, 0)
However, whilst running regression analysis I started to think that perhaps i should be treating the three groups (WLE, mastectomy, mastectomy & recon) as one variable and code them using dummy variables (especially as none of the patients who received a WLE had recon & therefore the previous "reconstruction" variables is influenced by what surgical procedure they underwent - i.e. no recon either means they had a WLE or mastectomy without recon).
However, reading up on creating & using dummy variables i have become confused as to how to do this, especially as i do not have a natural reference/control group.
Any ideas or advice would be greatly appreciated
This email was independently scanned for viruses by McAfee anti-virus software and none were found