曝光台 注意防骗
网曝天猫店富美金盛家居专营店坑蒙拐骗欺诈消费者
pre-determined categories of things that are of interest, some categories might
appear more interesting to the observer. If who, what or when to observe is left to the
whim of the observer – and to his or her interests – data recording may end up
focussing on some events to the exclusion of others. This is the essence of observer
bias, and is one of the largest problems in observational research. To reduce the
possibility of such bias, systematic sampling is used. The following details the
numerous types of sampling that are used in observational data collection.
3.8.1 Ad Libitum Sampling
This is equivalent to traditional field notes, and generally involves non-systematic,
informal observations preliminary to the quantified study. It is useful for recording
rare, unusual events, and for identifying relevant parameters in the preliminary
stages of an investigation. The “comments” column of a formal data-sheet can also be
used for this, but care should be taken to ensure such comments do not become the
focus of the research.
3.8.2 Continuous Sampling
In this method, all occurrences of specified behaviours and interactions are recorded.
This behaviour-change method usually records behaviour shown by a focal subject,
but can be modified to record focal behaviours, sequences, or locations. It also allows
for the calculation of frequencies and rates of behaviour. If behaviour termination or
transition times are also recorded, duration can be calculated. This method allows for
the most complete record of behaviour and is the only way to collect sequences
without missing anything. It can be very time-consuming and/or laborious if many
behaviours or subjects are involved, and the tempo is rapid. However, a sheet can be
designed to simplify data collection.
3.8.3 Instantaneous and Scan Sampling
These are time-sampling based systems in which the observer records the behaviour
state at the instant ending a predefined interval – for example, on the minute – which
is usually signalled by an auditory device heard only by the observer. By using such
a device, the observer is not having to constantly check the time, and can concentrate
on observing. To avoid bias, the observer only records what is going on at that point.
A potential problem is the difficulty in identifying the behaviour at a single glance.
Instead, it may require observing for 5 seconds to gauge what is occurring. The
observer can then record the behaviour at the end of the 5 seconds.
Instantaneous sampling is used where one subject is observed, while scan sampling
involves a group of subjects. In scan sampling, the observer must scan the group to
record the behaviour of all individuals. To avoid bias, it is important that even the
scan is systematic (for example, from left to right).
DSTO-TR-1034
11
Instantaneous and scan sampling approaches are the easiest ways of estimating the
percentage of time spent in specific activities. Thus, it is well suited to studies of
workload. It is less useful for data on interactions since they often occur in sequences
that cannot be recorded in a single scan. In addition, this method also misses
infrequent behaviours of short duration, unless the interval between samples is very
short or the entire observation is long. It is, however, a relatively simple method, and
naive observers can be quickly trained in its use.
The appropriate interval length depends on several factors, including activity level
(how often behaviour changes), group size (longer for more), or whether a single or
mixed sampling strategy is adopted (most common in trained observers). The shorter
the interval, the more the data represents that which would emerge with continuous
sampling. Shorter intervals also mean more data to analyse, but it is important in
rapid tempo operations. Longer scan intervals can be used during low tempo
situations. However, they need to be combined with some continuous sampling to
ensure that brief but important behaviours are recorded.
3.8.4 One-Zero Sampling
In one-zero sampling, time intervals are also used. However, each behaviour during
the interval is recorded as having an arbitrary score of 1, regardless of its actual
frequency. For example, a frequency of both 5 and 3 are recorded as “1”, while 0 is
entered if the behaviour was not observed. This does not allow the calculation of true
duration, rate or percentage, so it is not advisable for data collection where high
fidelity is important. It also over-estimates true percentage, when the results are
compared with continuous data. However, it is easy, and does produce high interobserver
reliability. Consequently, it may be the appropriate method to adopt when
中国航空网 www.aero.cn
航空翻译 www.aviation.cn
本文链接地址：
航空资料29(65)
