28
Stata Technical Bulletin
STB-4
Comments and a Policy Problem
What little machine manipulation is necessary with Stata can be insulated from beginning students with a system like this.
But do we really want to do this? And under what circumstances? Clearly, at one end of a continuum, we have classes that are
large, where the central theme of the class is not data analysis, in disciplines where the students’ quantitative and computer skills
are not well developed. Here a system like the one above can save instructors time and energy and under some circumstances
save the instructor from complete insanity.
On the other hand, where the class is small, where the central theme of the class is data analysis, where discipline or major
requires data analytic skills as prerequisite to the class or uses them extensively, what is the justification for a system which
keeps students from becoming familiar with the kinds of problems and situations they are likely to see if they attempt to practice
data analysis for their employer using the employer’s microcomputer and a purchased copy of Stata?
Anagnoson and DeLeon differ on this issue. DeLeon feels there is little justification for insulating students from the machine,
especially with microcomputers and Stata. Anagnoson definitely sees a need for insulation of students in the first situation above,
but feels that in the second situation, insulation is inappropriate.
Our system above is relatively primitive and easy to implement. One can go further and buy menu driven packages which
insulate students from the need to type commands. Two such packages are MicroCase and MIDAS. Since the latter uses Stata
as its statistical/data analysis “engine”, we have asked Professor Michael Macy of the Department of Sociology at Brandeis
University, the author of MIDAS and the author of several papers on the need for a new system of teaching statistics, for his
comments. They are to be found in tt2.
Other comments and suggestions are welcome. Send them to Joe Hilbe, STB Editor, and he will pass them on to us. Other
do- or ado-files that make Stata easier to use for students are welcome.
tt2 Using “front ends” for Stata
Michael Macy, Dept. of Sociology, Brandeis University
There is no general pedagogic principal that governs the “insulation” of students from statistical software and command
syntax. The use of an interface between the student and the command line depends entirely on the objectives of the instructor. If
the goal is to train students to use sophisticated stand-alone statistical software, to learn the mechanics of data management and
the mathematics behind computer routines, “friendly” programs like MIDAS, CHIP, or MicroCase may be inappropriate. Indeed,
when I teach introductory statistics, I have my students start out with MIDAS but I expect them to quickly move on to Stata.
However, many instructors, particularly those in the social sciences, tend to have different priorities. Their goal may be to
incorporate a laboratory component into a course that addresses issues for which data analysis may be a useful complement to
readings and lectures. Where the objective is to introduce liberal arts students to quantitative reasoning in an applied setting, the
use of stand-alone statistical packages may be counterproductive. Students are likely to be frustrated by the usual pitfalls that
await the novice researcher: arcane computer syntax, didactic and mathematical vocabulary, the disappearance of all their cases
through the conjunction of skip patterns, meaningless correlations between nominal variables, cross-tabulation of continuous
measures, etc. Rather than providing a highly motivating reinforcement for an introduction to quantitative reasoning, “hands-on
research” can become a burdensome and unsatisfying experience.
Faced with certain disaster, the instructor then has little choice but to develop lab exercises that simply walk the student
through the demonstration of the intended result—exercises that read rather like cookbook recipes, with no opportunity for
genuine inquiry. The problem is that students quickly grow tired of rote exercises and lockstep instructions. Once the novelty
wears off, most lab exercises tend to read like those dreadful instruction sheets included with children’s toys that carry the
warning label “Some assembly required.”
Dissatisfied with current lab curricula, I developed MIDAS as a “front end” for Stata. MIDAS consists of a command file
(written in PASCAL) and a Stata “do-file” that handshake with one another. The idea is essentially the same as Anagnoson
and DeLeon’s housekeeping programs, but I have simply pushed the principal a bit further. I wanted to not only simplify the
commands but to provide a structured research environment in which students could chart their own course without falling off
the edge of the earth. MIDAS does this by altering its menu-choices on the fly, depending on the user’s cumulative decisions. As
students gain confidence and savvy, MIDAS lets them “go behind” the interface and analyze their data directly with Stata, using
the command line. Indeed, that is my “hidden agenda!”
I suspect that readers who experiment with “housekeeping” do-files will end up doing the same thing I did as MIDAS
evolved...adding new features and routines. For those seeking a shortcut, I am happy to send copies of the MIDAS do-file to
readers who want to create their own “front ends” for Stata. Better yet, I will send you, on spec, the entire MIDAS program to
try.