Date: Tue, 4 Aug 1998 19:11:35 -0600
Reply-To: "Douglas J. Anderson" <Douglas.Anderson@ENMU.EDU>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: "Douglas J. Anderson" <Douglas.Anderson@ENMU.EDU>
Subject: Q-Q Normal Plots
Content-Type: text/plain; charset="us-ascii"
To Whom May Know The Answer:
I have been using the graduate pack edition for Power Macintosh of SPSS
6.1.1. I have been trying to do some data analysis for my masters thesis
research with the Normal Q-Q Plots. After doing quite a bit of reading in
the journals about these plots, I have several questions about the SPSS
method of obtaining them.
1) Why do the plots place the observed value on the X-axis and the
quantiles (often labeled the Qj from the expected normal distribution in
the literature) on the Y-axis? This is just the opposite of the standard
method of reporting the observed values on the Y-axis with the expected
normal quantiles on the X-axis. Every textbook and literature with graphs
on Q-Q plots places the observations the opposite of the SPSS method.
2) Why can I not swap axes so that my Q-Q graphs conform to the literature?
3) I would like to know whether there is a way that I can gain access to
the actual quantiles that the software must compute in order to plot the
expected normal values against the observed values of my data? I can
generate these values with a spreadsheet by long laborious calculations (I
am using the Blom method of (i - .385)/n +.25 with ties for rank broken by
the mean calculation). In this manner, I can then plot my data as a
scatterplot where I can control the X and Y axes and plot the data in the
customary manner. Also, I can then generate the correlation coefficient to
determine the best transform of my data, rather than rely on the
Shapiro-Wilk's statistic or K-S statistic produced by the Q-Q Plot of SPSS.
The problem is, it just takes too long to do this. Since the software is
calculating these quantiles, is there not some way that I can write a
syntax command to output this information to a file so that I can then use
it to produce my own scatterplots of observed against expected and generate
a correlation coefficient? (Please see Stephen W. Looney and Thomas R.
Gulledge, Jr., 1985, The American Statistician, February 1985, Vol. 39, No.
1 for the merits of this procedure.)
If there is anyone there reading this message who can help a graduate
student in anthropology/archaeology with this problem, then I would greatly
Douglas J. Anderson
Masters Candidate in Anthropology
Eastern New Mexico University
Portales, NM 88130
P.S. There is only one faculty member on our entire campus who uses SPSS
and he does not use or know anything about the Q-Q plots so I have no where
else to go for help.