Date: Mon, 14 Feb 2005 12:05:17 +1030
Reply-To: "Barnett, Adrian (HEALTH)" <adrian.barnett@HEALTH.SA.GOV.AU>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: "Barnett, Adrian (HEALTH)" <adrian.barnett@HEALTH.SA.GOV.AU>
Subject: Re: Comments on SPSS documentation
Content-Type: text/plain
Hi Dale (again)
I accidentally posted this to Dale alone, and am sending to the whole list
this time (sorry Dale! - thanks for your reply) as I'd like it to get a bit
more coverage - especially with people from SPSS.
Hi
I'd have to agree with you there. To some extent they have gone backward.
The older manuals were very enlightening on the statistical procedures used.
They were not a substitute for a text on the topic, but, given the wide
divergence in notation used by different texts, it was especially useful to
see how the procedure had actually been implemented. Picking up several
texts on a subject will not often enlighten you on where an how a particular
thing is in the output. Marija Norusis' manuals were excellent for this.
Even the old magenta-coloured manual from the 1970s was a pretty good intro
to a whole range of procedures.
Whilst I can see that it's not possible for SPSS to essentially write stats
texts they cannot just give a syntax for a procedure and leave you to work
out for yourself how it relates to what the textbooks cover.
In one area of syntax explanation they've gone backwards too. I've always
been baffled by the explanation of the third parameter on the INDEX command,
and consequently just ignored whatever it was trying to say. I puled out the
1983 edition of the SPSS-X User Guide and was amazed to find that there had
been a separate section which explained this and which has since been
removed form the manual. For the benefit of the list, I will reproduce it
below.
"7.18 The third argument of INDEX.
The third argument of INDEX is helpful when you need to look for more than
one character or set of characters in a string. For example, the expression
(INDEX,'MISSISSIPPI','LLSS',2) looks for either LL or SS. If the number of
characters in the second string is not evenly divisible by the third
argument, it is an error. If INDEX finds more than one string, it returns
the smallest index value. For example, the function
(INDEX,'MISSISSIPPI','PPSS',2) returns 3, not 9.
The most useful application of the third argument of INDEX is looking for
one among several special characters. For example, if two variables have
been recorded in two columns separated by either a blank, a comma, or a
semicolon, you could separate the variables by reading the data as a string
and separating it into numeric variables, as in:
DATA LIST NOTABLE /#VAR (A5)
COMPUTE #DELIM = INDEX(#VAR,' ,;',1)
COMPUTE VAR1 = NUMBER(SUBSTR(#VAR,1,#DELIM - 1),F2)
COMPUTE VAR2 = NUMBER(SUBSTR(#VAR,#DELIM + 1),F2)
PRINT /VAR1 VAR2 (F2,1X,F2)
BEGIN DATA
1 2
1,4
3,4
4,3
6;3
10 2
11;4
4 10
END DATA
#VAR is a scratch variable (Chapter6) used to read the first five columns of
data as a single alphanumeric variable, and #DELIM is a scratch variable
that stores the position of the delimiter character. In the COMPUTE command
for variable VAR1, SUBSTR(VAR,1,#DELIM - 1) says "Return the substring of
#VAR starting in the first position and ending one position before a blank,
comma or semicolon." Then SUBSTR(#VAR,#DELIM + 1) returns the substring
starting one position after any of these characters to the end of the
string. The two substrings are converted to numbers using the NUMBER
function (see Section 7.17). The PRINT display output is shown in Figure
7.18."
I think that without this bit, the explanation in the INDEX function of the
3rd parameter is entirely opaque. What a shame someone decided that 7.18 was
worth removing!!
The manual then goes on to explain that you can nest functions, a handy
technique that is no longer explained in the current manuals and which many
users would probably never work out for themselves. The explanation in the
section also provided valuable insights into how the different nested
functions worked, even in isolation. Another valuable piece of information
now lost.
I appreciate that the SPSS package is now much bigger than it was back then,
and requires much more stuff explained. However there is no longer an
argument based on problems of distributing many bound manuals, since this is
not done unless you buy them.
I think it is a pity that things which were once there, and were very useful
- and even indispensable, as in the above example - were discarded.
Regards
Adrian
--
Adrian Barnett
Research & Information Officer Ph: +61 8 82266615
Research, Analysis and Evaluation Fax: +61 8
82267088
Strategic Planning and Research Branch
Strategic Planning and Population Health Division
Department of Health email:
Barnett.Adrian@saugov.sa.gov.au
God is good but don't dance in a currach. - Irish proverb
This e-mail may contain confidential information, which also may be legally
privileged. Only the intended recipient(s) may access, use, distribute or
copy this e-mail. If this e-mail is received in error, please inform the
sender by return e-mail and delete the original. If there are doubts about
the validity of this message, please contact the sender by telephone. It is
the recipient's responsibility to check the e-mail and any attached files
for viruses.
-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
Dale Glaser
Sent: Sunday, 13 February 2005 5:49
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Comments on SPSS documentation
In the years of using SPSS since mainframe days, and the few years I have
been on this listserv, I have never used this as a forum for gripes, but I'd
like to register a minor one and see if I'm in the minority!! This past
week I had the opportunity to use PLUM for proportional odds modeling (AKA:
ordinal regression). When I accessed the manuals to aid in interpretation
of the Threshold and Location parameters, they are woefully amiss in
shedding any insight. Though I had a suspicion that the Threshold
parameters are calculated similarly to how the same parameters are
calculated via PRELIS for deriving the polychoric correlation matrix, I
wanted unequivocal confirmation of such, which the manuals do not do (though
I can always access the statistical algorithms if need be). That being
said, doesn't it seem in the "good old days" the SPSS manuals actually were
fairly illuminating in the mathematical/statistical details in the output?
I actually found them, alongside other texts, to be very educational.
Nowadays, the manuals are just on the CD and even then , they are nothing
more than details on point-and-click. Reading Long's "Regression Models for
Categorical and Limited Dependent Variables" last night made very clear the
interpretation for both the proportional odds model and multinomial logistic
model, but it would be nice to have that level of information in the SPSS
manuals.
Just a comment/frustration that I wonder if other SPSS users also
have........thank you...Dale
Glaser Consulting
Dale Glaser, Ph.D.
Consultant/Lecturer (SDSU/USD/AIU)
4003 Goldfinch St, Suite G
San Diego, CA 92103
phone: 619-220-0602
fax: 619-220-0412
email: glaserconsult@sbcglobal.net
website: www.glaserconsult.com
