I try to stress to my students the importance of screening their
data carefully for out-of range values, outliers, etc. prior to any other
analysis. I demonstrate to them how a single seriously out-of-range value
can ruin an analysis. I tell them that data screening is like washing the
dishes and cleaning the toilet -- nobody wants to do it, but if it is not
done, you are going to get sick. I hope I am getting through to them better
than did the instructors of my colleagues. Sometimes I wonder if it does
any good to harp on this -- maybe most people are just hopelessly
disinclined to attend to such boring details of their data.
Almost with fail, when one of my colleagues comes to me for help
with her analysis, I find serious data errors, even without setting out to
screen the data. This evening I found in a colleague's data a value of 26
on a variable that had a permissible range of 1-2. Not surprisingly, this
error had a dramatic and undesirable effect on the item statistics for that
item and the correlations of the affected scale with other variables. A
manuscript based on this erroneous analysis has already been submitted for
consideration for publication.
Another too frequent experience is misbehavior by SPSS. I rarely
encounter such problems with SAS. Twice today, with two different
variables, I asked SPSS to select cases that had, on the target variable,
values that were not equal to the system missing value for that variable.
Both times SPSS correctly excluded cases with missing values on the target
variable but also incorrectly excluded cases with values of zero on that
variable (zero was a permissible value).
As frustrating as dirty data and buggy software can be, they are
not nearly as aversive as a professional discussion group where most of the
posts are riddled with personal insults and where it is a rare post that
falls within the putative domain of the group -- the teaching of statistics.
----------------------------------------------------------------------------
----
Karl L. Wuensch, Department of Psychology,
East Carolina University, Greenville NC 27858-4353
Voice: 252-328-4102 Fax: 252-328-6283
[EMAIL PROTECTED]
http://core.ecu.edu/psyc/wuenschk/klw.htm
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
