Environmental Engineering Reference
In-Depth Information
the credibility of any study (see Fetterman, 1982b, for a detailed discussion of
this problem; see also Cook & Campbell, 1979).
Another problem with statistical tests is perceptual. Statistics demonstrate
correlations and not causality (unless the study—qualitative or quantitative—
is designed to ensure a proper temporal order of variables and controls for con-
founders). People frequently fall into the trap of inferring causality from
statistical correlation, however. Also, as Mark Twain wrote, “There are three
kinds of lies—lies, dammed lies, and statistics.” A competent individual can
manipulate figures to indicate almost anything. Statistical findings can be
mesmerizing for some individuals. Computer-generated findings are 10 times
as credible for no better reason than technological sophistication. Computer-
generated statistical errors can be particularly troublesome (if not materially
significant) because they are reproduced incorrectly throughout related data-
bases for long periods of time before detection. Countervailing forces or tests
cannot easily compensate for such systemic problems.
This brief review of problems should not dampen the spirits of an enter-
prising ethnographer. Ethnography has ample use for experimental designs,
quasi-experimental designs, and associated statistical analyses, including mul-
tiple regression analysis and factor analysis (Britan, 1978; Maxwell, Bashook,
& Sandlow, 1986). This brief review merely highlights the complexity that sta-
tistical analyses can create in ethnography or any other social science (for a
useful presentation of statistics for the social sciences, see Blalock, 1979;
Hopkins & Glass, 1987; Hopkins, Hopkins, & Glass, 1995; Shavelson, 1996).
CRYSTALLIZATION
Ethnographers crystallize their thoughts at various stages throughout an ethno-
graphic endeavor. The crystallization may bring a mundane conclusion, a
novel insight, or an earth-shattering epiphany. The crystallization is typically
the result of a convergence of similarities that spontaneously strike the ethnog-
rapher as relevant or important to the study. Crystallization may be an exciting
process or the result of painstaking, boring, methodical work. This research
gestalt requires attention to all pertinent variables in an equation. Gross errors
can be misleading and lethal to any investigation. For example, a long line of
cars with their lights on, all proceeding down the same street in the same direc-
tion, might suggest a funeral procession. However logical this conclusion
might appear, it may also be dead wrong. Additional data from informal inter-
views or more disciplined and detailed observations are necessary. For
example, identification of a hearse in the procession or a confirming word from
one of the individuals involved would immeasurably improve the credibility of
