Biomedical Engineering Reference
In-Depth Information
searches for published and unpublished studies of the appropriate study
design; critical appraisal of the internal and external validity of these
studies; extraction of relevant data about their methods and results; assess-
ment of study heterogeneity and, where appropriate, meta analysis (qv.) or
other methods to synthesize their results.
Tasks:
Test cases against which the performance of human participants or
an information resource is studied.
Triangulation:
Drawing a conclusion from multiple sources of data that
address the same issue. A method used widely in subjectivist research.
Two by two (2
2) table:
Contingency table (qv.) in which only two vari-
ables, each with two levels, are classified.
¥
Type I error:
In statistical inference, a type I error occurs when an investi-
gator incorrectly rejects the null hypothesis, typically inferring that a study
result is positive when it is in fact negative.
Type II error:
In statistical inference, a type II error occurs when an inves-
tigator incorrectly fails to reject the null hypothesis, typically inferring that
a study result is negative when it is in fact positive.
Usability testing:
A type of pilot study (qv.) in which the focus is on eval-
uating the ease of use of the resource. (See www.useit.com).
Validation:
(1) In software engineering: the process of determining
whether software is having the intended effects (similar to evaluation). (2)
In measurement: the process of determining whether an instrument is mea-
suring what it is designed to measure. (See Validity).
Validity:
(1) In demonstration studies or experimental designs: internal
validity is the extent to which a study is free from design biases that threaten
the interpretation of the results; external validity is the extent to which the
results of the experiment generalize beyond the setting in which the study
was conducted. (2) In measurement: the extent to which a instrument mea-
sures what it is intended to measure. Validity is of three basic kinds: content,
criterion-related, and construct. (See Chapter 5).
Variable:
Quantity measured in a study. Variables can be measured at the
nominal, ordinal, interval, or ratio levels (qv.).
Verification:
Process of determining whether software is performing as it
was designed to perform (i.e., according to the specification).
