Biomedical Engineering Reference
In-Depth Information
influence, but it can be used as evidence that the influence is
less than the predicted amount. “Proof,” of course,
and are reported as a frequency, or frequency is divided by
observation time to calculate a rate, like slaps per hour.
Other types of responses, such as sitting in contact, are
behavioral states, and measuring the frequency of these
responses would be difficult. The durations of bouts of
sitting in contact vary; bouts can be long and may be briefly
interrupted as individuals adjust their positions. In this
regard, a duration measure may be more useful than
a frequency measure is, and may be expressed as minutes
per hour in contact or percent of time in contact or any
similar measure.
Some acts have discrete times of onset and cessation
and are relatively independent of one another. Questions
concerning how long the average chase lasts in a group can
be answered by recording both the frequency and duration
of the chase and dividing the latter by the former. On the
other hand, some acts are brief but are clustered in bouts;
that is, each act is not independent of the occurrence of the
previous act. Individuals engaged in play frequently pause
and resume the activity; therefore, the frequency of play is
hard to define and the duration can either include or exclude
the brief pauses. For example, a pair of individuals may
slap each other repeatedly in little bursts or bouts during
a fight. The fights may vary in duration, and the number of
slaps delivered may vary from fight to fight. The duration of
slapping may thus be impossible to assess and may be
meaningless as a measure of aggressive behavior. The
frequency of slaps may also be a meaningless measure of
aggressive behavior. In such a case, the frequency and
duration of fights,
is
a probability statement. A probability of
0.01 means that
there is a 99% chance that these results did not occur by
chance. Normal distributions imply that all events can
happen but that some events are very rare. For example,
four-leaf clovers may be rare, but if one looks through
enough clovers, he or she will find several four-leaf clovers.
The probability of finding a four-leaf clover in a collection
of four-leaf clovers is very high.
Experimental methods ultimately rely on measurement
and definition. Conclusions are probability statements, and
tests require quantitative data, even if only on a yes/no
scale. An independent variable may be manipulated by its
presence or absence, e.g. a diet with an additive or a diet
without an additive. An independent variable may be
manipulated on a graded scale, as in grams per kilogram of
diet or by percent of diet or by absolute amount. Indepen-
dent variables can be useful even when they cannot be
directly manipulated as long as they can be measured.
Experiments can be performed when the independent
variable is morning or afternoon, winter or summer, or pre-
or post-pubertal period. These times are not manipulated
but can be measured and assigned to subjects as
a treatment.
When using an independent variable, like season, to see
if a dependent variable, like aggressive behavior, varies
with the season, one might be tempted to measure all
possible behavioral differences as a function of season, that
is, the influence of a host of independent variables on
a dependent variable, like aggressive behavior, or the
influence of a single independent variable, like season, on
a host of dependent variables (like 100 different ethogram
items). As the number of measured variables increases, so
does the possibility that a difference will be found because
of a sampling error. If enough coins are tossed 10 times
each, one eventually will show 10 heads in a row even
though none of the coins is truly biased.
<
rather
than the slaps, should be
measured.
Since fights take on various forms and often lack
a single motor description, the objective is to operationally
define a “fight.” One possibility for achieving this is to
measure probability rather than frequency or duration of
the event. Using this option, one could measure the prob-
ability of seeing one or more slaps if the subject is observed
for n minutes. If the probability is higher in one group than
in another, then one might conclude that slapping is more
likely to be seen in that group than in the other group,
despite not knowing the true frequencies and durations
involved.
Behavioral Data Collection
Data can be measured under a variety of circumstances and
by multiple techniques. A sample should represent the
population from which it was drawn. It is seldom possible
to measure every individual in a population; however, the
goal is to obtain an unbiased representative sample. Data
collection can then proceed under the specified conditions.
When collecting behavioral data, one may focus on acts
or responses that are essentially instantaneous. For
example, a slap may have a finite duration, but that duration
is very brief and the amount of variation in the duration of
slaps is very small. If slaps are distributed in time randomly
without respect to one another, then each slap is an inde-
pendent measure of slapping. These data are distinct acts
One
zero Sampling
Altmann (1974) reviewed all the data collection schemes
commonly used to measure frequency and duration.
A common technique used at the time of her review was
one
e
zero sampling, which recorded a 1 if a response
occurred with a set time period or a 0 if it did not. Altmann
argued that one
e
zero sampling provides data on neither
frequency nor duration but an indeterminate index of both.
She dismissed one
e
zero sampling as useless for measuring
frequency and duration but suggested using this technique
e
Search WWH ::




Custom Search