Finding Your Way Through the Forest - Ethnography - page 107

determine if there was a difference. The results were statistically significant at
the p = 0.04 level. (An analysis of the same courses, but removing the one
course that did not engage in the process, resulted in highly statistically sig-
nificant findings at the p = 0.01 level.) This statistical measure independently
reinforced the qualitative ways of knowing that focused on the level of com-
munity engagement, participation, and intellectual excitement.
Ethnographers also use the results of parametric statistics, as well as test
scores, to test certain hypotheses, cross-check their own observations, and
generally provide additional insight. Student test scores were essential to one
portion of the CIP study. The sponsors wanted to know if the students' reading
and math capabilities improved as a result of their participation in the program.
Gains in reading scores were statistically significant. From the sponsors' and
the ethnographer's perspective, this information was a useful finding. The
gains in math scores were statistically significant but less spectacular than the
gains in reading scores. This particular finding provided the ethnographer a
unique opportunity to interact with the psychometrician in a significant—
interpretive—fashion. The statistical calculation delivered an outcome but not
the process behind it. Ethnographic description was useful in explaining why
the math gains were not as spectacular as those in reading. The answer was
simple: The math teaching positions were vacant during most of the study. The
program had difficulty recruiting and maintaining math teachers given the
competitive market for these individuals.
The test outcomes were a product of traditional psychometric approaches,
including control and comparison group data using analysis of covariance and
standardized gain procedures. This information was both useful to sponsors
and the ethnographer and valuable in providing a focal point for further
inquiry and data comparison.
In addition to commercial statistics packages, such as SPSS, I use online
statistical programs. One of the strengths of these programs is their ability
to graphically demonstrate probability distributions. They also have spread-
sheets, facilitating sorting, arithmetic and mathematical transformations
(including z or N scores), statistical tests including descriptive statistics, con-
fidence intervals, paired and independent comparisons (parametric and non-
parametric), correlations, linear regression, and contingency tables. They also
generate scatterplots, histograms, boxplots, stem and leaf diagrams, and pie
charts. They typically include both static probability tables (e.g., normal, t and
F distribution, and chi-square distribution) and active tables. (See Handwerker,
2001, p. 222, for examples of parametric statistics ranging from factor analy-
sis to logistic regression. See also Boruch, Weisburd, Turner, Karpyn, &
Littell, 2009, and Mark & Reichardt, 2009, concerning experimental and
quasi-experimental design.)