Geoscience Reference
In-Depth Information
2.1 Statistical treatment
Simple random sampling was adopted and the size of the sample (269
users) was developed in order to satisfy the condition of 90% confidence
level, a study proportion of 50% and a margin of error of 5% (WITTE and
WITTE 2005). Three experimental groups were determined: choropleth
(chor), dot (dot) and range-graded proportional symbol (prop) users. Dif-
ferences between correct and wrong answers and among these groups were
analyzed by a Kruskal-Wallis ANOVA test, at 95% confidence level,
which can determine if there is a significant variation due to the technique
used acting like a common analysis of variance. This test is used when
there is no evidence of a normal distribution. Also, under the analysis of
variance, a multiple comparison procedure was applied in order to identify
where the differences pointed out by the Kruskal-Wallis occur, using
Fisher's least significant difference (LSD) procedure.
3- Results and discussion
Students who agreed to participate in the test sample seemed to be heavily
involved in the subject according to the amount of written comments about
questions and maps in the appropriate spaces in the questionnaire. This
seemed to ensure the validity of the methodological approach and con-
sequently of the analyzed answers. Most of decisions made by respondents
were well explained and had logical reasoning. Errors mainly seemed to
occur because of lack of attention and consequently missed outliers,
instead of random guessing or disinterest in the questions. Talking with
students after the test revealed that they considered the questionnaire easy
to answer and the instructions clearly given, and almost no questions about
this were registered in the 11 classes in which this test took place, between
July and August, 2010.
The two main general performance scores show details about questions
that are part of the measurement of effectiveness for the two tasks
proposed for this work. For the direct acquisition task ( Figure 4 ) , the mean
accuracy was about 91.5%. The question with the worst results was 'o14':
“At least eight parishes did not reduce “child mortality” to zero in 2009”.
For the map reasoning task ( Figure 5 ), six from seven of the most cited
were critical (most severe) and urgent situation parishes. The 'p15' question
Search WWH ::




Custom Search