Reliability is an assessment instrument, the tool in measuring what is measured. This means that whenever the instruments used will yield the same measure. The most obvious example is the scale or meter. The same thing happened to gauge a symptom, behavior, traits or characteristics of individuals and others. For example gauge learning achievement such as achievement test, attitude measuring instruments, questionnaires, etc., should examine the nature of such constancy.
Achievement test is said to remain if the current measurement results showed similar results at a different time, against the same student. For example, students in grade V on this day in tests of mathematical ability. The next week students in the test back. Results from both tests are relatively similar. Nevertheless is still possible there are differences in results for certain things due to accidental factors, time interval, the change students' views on the same matter. If this happens, the weakness lies in the measuring instrument, which has no certainty or doubt the students' answers. In other words the degree of reliability is still low.
On the other hand, differences in measurement results is not caused by the measuring tool, but a condition that happens to students. For example, students in a state of physical pain during the first test, the motivation at the time of the first test is different motivations on the next test.
On the basis of the difference the first measurement results with the results of subsequent measurements can teijadi due to change in subjects who were measured and / or by factors relating to the granting of the test itself. This is not surprising and are common, often expressed with the title / term measurement error. This means, the score of the first measurement results and measurement results of the second score on the same subject, possible measurement errors caused by two factors above. Thus each measurement score results produce two parts, namely the results of the first measurement, called the true score and the results of subsequent measurements on the same subject, which contains the results of the score plus measurement error.
True score and score components that contain errors of measurement is expressed in a mathematical equation as follows:
with:
X = Observed score
b = true score
s = measurement error
In a study score observed was a true score plus measurement error score so that the observed score variance X2 is Tb2 true score variance plus error score variance or Tx2 = Tb2 TS2 TS2.
Reliability index measuring instrument in a study can be found by correlating the scores obtained from repeated measurements at different times, or with a group of questions is worth it. This procedure is done by giving the test twice to the same subject at different times. The second way is to divide the measuring instrument (test) into two equal parts or the Setarap to view the constancy test. The first way is known to re-test (test retest) and the second way is known as fractional comparable / equivalent.
a. Retest reliability
Retest (test-retest) is the use of measuring instruments on the subject are measured, made twice in different times. For example, mathematics achievement test for fifth grade elementary school students, given today, then checked the results. A week later the test was given again on the same students and the results are examined. The results of the first measurement and then correlated with the results of the second measurement to obtain the correlation coefficient (r). The correlation coefficient is called coefficient of retest reliability, the results will move from - 1.0 to 1.0. When the reliability coefficient close to 1.0 is a high reliability index. This means that the first measurement results are relatively similar to the results of the second measurement. In other words, the instruments have a level of constancy or provisions of (reliable). To measure the social sciences and education reliability index 0.75 was considered sufficient given the nature and the social sciences and education are different from exact sciences.
The distance or interval between the first measurement with the second measurement should not be too close nor too far. If too close / short, the measurement results are influenced by memories of the students on the answers given at the first measurement, not because the measuring instrument constancy. Conversely, if the first measurement time interval with the second measurement is too long, there could be a change in students' knowledge and experience that affect the reliability coefficient. The assumptions used in the retest is characteristic measured by the measuring instrument is stable over time, so if there is a change score of the two measurements is due to measurement error. How to re-test (test-retest) are widely used in establishing or determining the level of reliability of measurement in social research and education.
b. Reliability equivalent fractions
Reliability is not done a fraction equal to the subject repeated the same measurements but using results from tests of a comparable or equivalent to that given to the same subject at the same time. Thus the required two sets of measuring instruments are arranged in such a way as to have a degree of similarity or equality in terms of, the content, level of difficulty of measuring instruments, abilitas measured, the number of questions, the form of questions and other technical aspects. The difference is only a question. When the constituent instrument of equality can be achieved optimally then the coefficient of reliability of this procedure is considered the most well compared with the retest procedure. But the difficulty lies in setting measurement devices that actually contain a high degree of equality.
Achievement test is said to remain if the current measurement results showed similar results at a different time, against the same student. For example, students in grade V on this day in tests of mathematical ability. The next week students in the test back. Results from both tests are relatively similar. Nevertheless is still possible there are differences in results for certain things due to accidental factors, time interval, the change students' views on the same matter. If this happens, the weakness lies in the measuring instrument, which has no certainty or doubt the students' answers. In other words the degree of reliability is still low.
On the other hand, differences in measurement results is not caused by the measuring tool, but a condition that happens to students. For example, students in a state of physical pain during the first test, the motivation at the time of the first test is different motivations on the next test.
On the basis of the difference the first measurement results with the results of subsequent measurements can teijadi due to change in subjects who were measured and / or by factors relating to the granting of the test itself. This is not surprising and are common, often expressed with the title / term measurement error. This means, the score of the first measurement results and measurement results of the second score on the same subject, possible measurement errors caused by two factors above. Thus each measurement score results produce two parts, namely the results of the first measurement, called the true score and the results of subsequent measurements on the same subject, which contains the results of the score plus measurement error.
True score and score components that contain errors of measurement is expressed in a mathematical equation as follows:
X = b s,
with:
X = Observed score
b = true score
s = measurement error
In a study score observed was a true score plus measurement error score so that the observed score variance X2 is Tb2 true score variance plus error score variance or Tx2 = Tb2 TS2 TS2.
Reliability index measuring instrument in a study can be found by correlating the scores obtained from repeated measurements at different times, or with a group of questions is worth it. This procedure is done by giving the test twice to the same subject at different times. The second way is to divide the measuring instrument (test) into two equal parts or the Setarap to view the constancy test. The first way is known to re-test (test retest) and the second way is known as fractional comparable / equivalent.
a. Retest reliability
Retest (test-retest) is the use of measuring instruments on the subject are measured, made twice in different times. For example, mathematics achievement test for fifth grade elementary school students, given today, then checked the results. A week later the test was given again on the same students and the results are examined. The results of the first measurement and then correlated with the results of the second measurement to obtain the correlation coefficient (r). The correlation coefficient is called coefficient of retest reliability, the results will move from - 1.0 to 1.0. When the reliability coefficient close to 1.0 is a high reliability index. This means that the first measurement results are relatively similar to the results of the second measurement. In other words, the instruments have a level of constancy or provisions of (reliable). To measure the social sciences and education reliability index 0.75 was considered sufficient given the nature and the social sciences and education are different from exact sciences.
The distance or interval between the first measurement with the second measurement should not be too close nor too far. If too close / short, the measurement results are influenced by memories of the students on the answers given at the first measurement, not because the measuring instrument constancy. Conversely, if the first measurement time interval with the second measurement is too long, there could be a change in students' knowledge and experience that affect the reliability coefficient. The assumptions used in the retest is characteristic measured by the measuring instrument is stable over time, so if there is a change score of the two measurements is due to measurement error. How to re-test (test-retest) are widely used in establishing or determining the level of reliability of measurement in social research and education.
b. Reliability equivalent fractions
Reliability is not done a fraction equal to the subject repeated the same measurements but using results from tests of a comparable or equivalent to that given to the same subject at the same time. Thus the required two sets of measuring instruments are arranged in such a way as to have a degree of similarity or equality in terms of, the content, level of difficulty of measuring instruments, abilitas measured, the number of questions, the form of questions and other technical aspects. The difference is only a question. When the constituent instrument of equality can be achieved optimally then the coefficient of reliability of this procedure is considered the most well compared with the retest procedure. But the difficulty lies in setting measurement devices that actually contain a high degree of equality.
c. Reliability halved
Reliability split similar to the reliability of equivalent fractions, especially from its implementation. In this procedure measuring instrument given to the subject just once or one time. Beads of questions divided into two comparable parts, usually distinguish questions about the even number with an odd number. Every part about checked the results, then scores from the two sections correlated to look for correlation coefficients. Given the correlation was only half do not apply to the whole question, then the correlation coefficient was not acquired for the entire problem, but only half. Therefore, the correlation coefficient halved need to be changed into a correlation coefficient for the entire problem by using Spearmen Brown prophecy formula. After doing the conversion. Assumptions used in the split second procedure is the second part of the parallel measuring instrument, though often wrong or incorrect. As a result of the conversion coefficient of reliability, the split second procedure tends to show high reliability coefficients than the re-test procedure and equivalent fractions. Therefore, the use of the split second should be more careful. This procedure is used when measuring devices containing or consisting of many items, the item is relatively heavy / hard (power test), the tested material comprehensive enough to enable the preparation of two questions to one similar problems to meet the split second.
d. Similarity rational
In addition to the ways described above there are procedures for calculating reliability without performing measurements or correlation of two equivalent fractions and the split second. The way it is rational similarity. This procedure is done by connecting each point in one test with other grains and with the test itself as a whole. One way is to use a formula often used Kuder-Rechardson or KR 21.
Reliability split similar to the reliability of equivalent fractions, especially from its implementation. In this procedure measuring instrument given to the subject just once or one time. Beads of questions divided into two comparable parts, usually distinguish questions about the even number with an odd number. Every part about checked the results, then scores from the two sections correlated to look for correlation coefficients. Given the correlation was only half do not apply to the whole question, then the correlation coefficient was not acquired for the entire problem, but only half. Therefore, the correlation coefficient halved need to be changed into a correlation coefficient for the entire problem by using Spearmen Brown prophecy formula. After doing the conversion. Assumptions used in the split second procedure is the second part of the parallel measuring instrument, though often wrong or incorrect. As a result of the conversion coefficient of reliability, the split second procedure tends to show high reliability coefficients than the re-test procedure and equivalent fractions. Therefore, the use of the split second should be more careful. This procedure is used when measuring devices containing or consisting of many items, the item is relatively heavy / hard (power test), the tested material comprehensive enough to enable the preparation of two questions to one similar problems to meet the split second.
d. Similarity rational
In addition to the ways described above there are procedures for calculating reliability without performing measurements or correlation of two equivalent fractions and the split second. The way it is rational similarity. This procedure is done by connecting each point in one test with other grains and with the test itself as a whole. One way is to use a formula often used Kuder-Rechardson or KR 21.
No comments:
Post a Comment