Computer
experience history Jim Groeneveld
Since about the
early 70-ies I have been using computers, initially a mainframe computer (CDC
Cyber with OS SCOPE, later NOS and NOS/VE) with (as far as I remember) a memory
of 100,000 words (1 word = 10 bytes of 6 bits each). It was driven by Hollerith
punch cards (80 columns), punched on very noisy punch machines. From there we
punched both our programs and our data, which we delivered at the appropriate
desk. After an inevitable ‑coffee‑ break of at least 15 minutes,
often more, we could soon fetch the output from a shelf, study the errors and
rerun the whole process, this time only repunching
the buggy cards. After several iterations we were mostly satisfied enough with
the result. I still can hear those punching machines in my mind.
Some time later we
applied TTY's (printing terminals, 110 bps) instead of punch cards,
or even real electronic terminals (300 and 1200 bps) via direct or modem (telephone)
lines. That way we could remain at our working place, quite a distance away,
enter our data, compose our program, run the whole combination, preview the
results on the screen, and having it printed at the computer centre. We
received the printed output (with line numbers) via regular mail the next
morning to study and revise. Each day an iteration like that took place, trying
to do as much as surveyable from the output, the
changes written in it with red ink, and the scrolling line oriented editor. This
really already was a luxioury.
At that time we
used a locally (at the State University of Groningen) written statistical
package, called WESP, in Dutch: "Waarlijk Eenvoudig Statistisch Pakket", in English: "Really Simple Statistical
Package". I don't remember much of it anymore, but variables didn't have
names, but just sequence numbers. Data editing was limited, and so was the
amount of statistics, though at that time it was quite worthwhile and offering
almost every we needed for our projects.
During the late
70-ies we also started using so called table calculation machines, programmable
calculators actually of the size of a typing machine, from Diehl. They could be
programmed in their own, very specific machine language. I remember the Combitronic with 100 programming steps and 10 numerical
registers, for which I a.o. programmed the binomial
distribution outcomes for given numbers and proportions.A
quite sophisticated one was the Alphatronic, having a
memory of 1600 programming steps, equivalent to its 160 numerical registers,
which were mostly divided into 800 steps with 80 registers. I programmed a
2-way analysis-of variance program on it, also to be used as 1-way with
repeated measures. The allowed cell dimensions were 10 by infinite (really infinite,
as running calculations took place) and supported designs were both equal and
unequal cell frequencies with several statistical variations (after Winer) and homogenity-of-variance-tests
(Hartley and Cochran, I believe) as a bonus. I still must have the cassette
with the software on it somewhere.
Gradually I made
acquaintance with SPSS-6000 (vs. 6 up to 9) from
Since computers
could print texts I have been applying text processing from its most elementary
shape. At a certain stage we had access to a real text processing package on a
mainframe. It, however, could not automatically hyphenate words. So, in the
early 80-ies, I wrote my own fully automatic (Dutch) hyphenation software
(KAPAF) in Fortran, which was based on fuzzy logic,
hyphenating probabilities between the characters of a pair of letters. It
functioned surprisingly well as a preprocessor with the text processing
software, and has been used for many years while producing our study reports.
SPSS changed its
syntax drastically around the middle of the 80-ies, with new SPSS-X versions
for mainframes (which also ran on a VAX/VMS, that I used much later) and
SPSS-PC for the IBM-PC. Around 1991 years ago I wrote an almost fully automatic
syntax converter in BASIC (SPSS9TOX) for the IBM-PC to use with old ascii data files, which could not be reread and managed
differently (i.e. converted to the new SPSS system file format). Since 1985 I
have been using PC's or compatibles (running DOS) and
from around 1990 these remained the only computers to use. Our graphical
applications were either plots from SPSS or consisted of very nice graphical
screens and prints from Statgraphics (vs. 2.x). All
the time I have been busy with data conversion quite a bit (including the
necessary -serial- communication) and wrote a.o. the
program DATAFIX (initially in BASIC, later in rewritten C) in particular to
manage very long records (line lengths in ascii
data files).
During the early days
(first ten years) of the IBM-PC and its successors much smaller or larger
freeware statistical software was developed in the world. At the end of the
80-ies I had them all in control and presented a comprehensive review on them.
Later statistical software developments were fast and my time was limited, so I
lost control.
During the early
90-ies I was involved in a (medical informatics) project developing
client-server software (written in C) with a very user friendly GUI for the
HP-Unix platform, driving all kinds of Unix or DOS
programs in the background. My job was to build the statistical client and
server, i.e. to build a GUI, interactively presenting all kinds of statistical
techniques to be run from the server. The server part consisted of a computer
language generator, adaptable to any statistical package with an appropriate
configuration file, e.g. EpiInfo and SPSS. Currently (2014) I am gradually switching from
WXP to Ubuntu Linux.
Only from 1997 I
started to learn and use SAS (vs. 6.12) for Windoesn't.
I have the impression, that SPSS is slower in comparison with SAS, while
performing similar tasks. This may be due to the fact that SAS compiles its
DATA and PROC steps before running, while SPSS has always remained an
interpreter (as far as I know). Since 1999 I am active on the internet
discussion list SAS-L as a SAS expert and have been nominated as
Rookie-of-the-year 1999. In the meantime I passed my SAS Base, Advanced and
Clinical Trials Certified Professional V9 exams.
Well, actually I
have been studying educational psychology (and graduated in that), but
specialized in social scientific research, its methodology, statistics and
computer programming and I have been employed in research and data processing
all the time.
Last update: 6 August 2014