Reliability is a characteristic of a methodology that reflects the accuracy of
psychodiagnostic measurements, as well as the stability of test results to the action of
extraneous random factors. The results of a psychological study are usually influenced by a
large number of factors (emotional state and fatigue, if they are not included in the range of
characteristics under study; physical factors: illumination, temperature and other features of
the room where the study is conducted; the level of motivation
of the subjects for the
examination; the influence of the personality of the experimenter). Any change in the
research situation enhances the influence of some and weakens the influence of other
factors on the test result.
apparatus (first of all, the correlation method and factor analysis) to substantiate the degree
of compliance of the methods with the noted criteria.
estimates.
The distribution of test scores when performing a test that measures one quality ideally
coincides with a normal distribution, and the variance will be “true” (i.e., reflecting the
variability of only the measured trait). Each subject occupies a certain place according to
the test scores, and theoretically this place is constant for each member of the sample. In
the case under consideration, repeated performance of the test by the same persons should
give a distribution of places on the rating scale identical to the first one. Then the technique
as a measuring
tool is accurate and
method reliability.
Normal distribution is a kind of theoretical distribution of variables. It is observed when
measuring a sign (variable) under the influence of many relatively independent factors.
Extreme values
As A. Anastasi writes, it is hardly possible to trust the intelligence test if at the beginning
of the week the child had an indicator equal to 110, and by the end - 80. Repeated
application of reliable methods gives similar results.
In psychodiagnostics, the problems of reliability and validity of methods are closely
interrelated, however, there is a tradition of separate presentation of these most important
characteristics. Let's start by looking at it
The reliability of a test in a broad sense is a measure of the extent to which the
differences in test results found in a number of subjects are a reflection
of actual differences
and to what extent they can be attributed to random errors.
The reliability of a test in the narrow sense is the degree of consistency of test results
obtained during its initial and repeated application in relation to the same subjects at different
times using different, but comparable in nature, sets of test items.
most reliable.
The overall scatter (dispersion) of the test results can thus be represented as the
result of the influence of two groups of causes: the variability inherent in the measured
property itself, and the factors of instability of the measurement procedure.
Machine Translated by Google
and the direction of the relationship between the two variables. The lack of
connection is expressed by values close to zero.
are quite rare in it; values close to the average -
often enough.
In
practice, three main methods for assessing reliability are used.
tests: retesting, parallel testing and splitting.
With an increase in the time interval between tests, the
correlation
indicators decrease, so the interval should not exceed 1-3 months. This method
of determining reliability has a drawback. With the repeated application of the
same tasks and, especially after a short time, the skill of working with the
methodology is developed and, consequently,
the individual indicator of
reliability worsens.
2. Parallel testing (reliability of parallel forms) is determined
using the interchangeability of test forms in two ways:
If we deviate from the ideal case, then the real assessments and ranking
places of the subjects change during the second examination, and their
distribution differs to some extent from the initial one. In this case, the variance
of the new distribution is higher than the original one by the value of the
variance of the measurement error. Measurement error is a statistic that
characterizes the degree of accuracy of individual measurements. The reliability
of the test is closely related to the measurement error, which indicates the
likely limits of fluctuations in the measured value under the influence of random constant factors.
1) the same test samples are examined first
using the main set of tasks,
after some time using a similar but additional set;
1. Retesting (retest reliability) -
2) the subjects are divided into equal groups, one of the groups is tested according
to the form of the A-test (main), and the second - according to the form of the B-test
(additional). In two weeks, on the contrary, the test forms for groups change.
In practice, in most of the methods used, it is rarely possible to obtain a
value of reliability factors exceeding 0.7-0.8. The technique is recognized as
reliable when the obtained coefficient is not lower than +0.75 - +0.85. The best
reliability tests give coefficients of the order of +0.90 or more. (The correlation
coefficient is a normalized quantitative indicator that varies from -1.00 to +1.00
and evaluates the strength
The main indicator here is the correlation coefficient between the results
of the initial and repeated examinations, which evaluates both the temporal
stability of the test (actual reliability) and the degree of agreement between
the results of both forms of the test.
a characteristic obtained by re-examining subjects using the same test over a
time interval. Reliability in this case is calculated by the correspondence
between the results of the first and second surveys or by maintaining the rank
places of the subjects in the sample during the retest. The reliability coefficient
corresponds to the coefficient of correlation between the results of such surveys.
Machine Translated by Google
Internal consistency (consistency) - type of test reliability - homogeneity of
a
set of test items, i.e. the degree to which particular items evaluate the same
construct, how much they measure the same variable. One type of internal
consistency assessment is splitting reliability.
The relationship between the parallel forms of the test is complex. Both
sets should not only meet the same requirements, measuring identical
indicators and giving similar results, but, at the same time, should be relatively
independent of each other. In practice, this task is not feasible for all test tasks
(especially for personal methods, questionnaires), which significantly limits the
scope of this type of reliability.
3. Splitting (reliability of parts of the test) -
reliability characteristics
obtained by analyzing the stability of the results of individual sets of test tasks
or single items of test items.
Do'stlaringiz bilan baham: 
