Reliability, Validity, Accuracy & Errors!
Reliability, Validity & Accuracy are concepts that must be analysed when reporting scientific results. When considered collectively, they provide a strong assessment of instruments and/or the entire scientific investigation method. When discussing reliability, validity and accuracy, it is important to identify problems, how they were minimised or how they can be minimised for future investigations - this demonstrates critical thinking.
Reliability:
As defined by the NSW Board of Studies, reliability is the degree with which observation and/or measurements taken under identical circumstances will yield the same results.
Repeating the experimental method at least 3 times ensures reliability (in exams, say 5!). To effectively evaluate reliability, both the processes and data analysed. Ultimately, the conditions under which data is measured must be identical, thereby yielding similar results. The following questions should be evaluated in a discussion of reliability:
Validity:
As defined by the NSW Board of Studies, validity is the extent to which the processes and resultant data measure what was intended.
As such, gathered data must be completely relevant to the aim and hypothesis of the investigation. Evaluation of validity involves analysis of both the processes and data obtained. The following questions should be evaluated in a discussion of validity:
Accuracy:
Accuracy is the closeness of a measured value to its true values and allows valid conclusions to be made. The accuracy of data is considered by analysing the precision of measurement and errors. One way to enhance the accuracy of data is to use measuring instruments with relatively a relatively small limit of reading. Errors in experimental design harm the accuracy of data and ultimately impede both the reliability and validity of the investigation. Several causes of inaccurate results are:
Random Errors:
Random errors are caused by unknown and unpredictable changes in the experiment (e.g. due to the instruments or environmental conditions). These are statistical fluctuations in both directions about the true value and thus repetition and statistical analysis can reduce such effects. Random errors can be reduced in several ways:
Systematic Errors:
Systematic errors are caused by measuring instruments either being used incorrectly or possessing damage to their structure and thus harming their function. Systematic errors limit accuracy and thus the following steps should be taken to reduce such errors:
Reliability:
As defined by the NSW Board of Studies, reliability is the degree with which observation and/or measurements taken under identical circumstances will yield the same results.
Repeating the experimental method at least 3 times ensures reliability (in exams, say 5!). To effectively evaluate reliability, both the processes and data analysed. Ultimately, the conditions under which data is measured must be identical, thereby yielding similar results. The following questions should be evaluated in a discussion of reliability:
- Has at least 4 trials been conducted?
- Has quantitative data been averaged?
- Have the same instruments/observers been consistently used for each trial?
- Have random errors (such as parallax error, human judgement/reflex error) been minimised?
- Have instruments with a suitable limit of reading been used (generally, the smaller the limit of reading, the greater the reliability)?
Validity:
As defined by the NSW Board of Studies, validity is the extent to which the processes and resultant data measure what was intended.
As such, gathered data must be completely relevant to the aim and hypothesis of the investigation. Evaluation of validity involves analysis of both the processes and data obtained. The following questions should be evaluated in a discussion of validity:
- Was the appropriate independent and dependent variables selected?
- Was the selected measuring apparatus used able to measure the intended independent and dependent quantity(s) accurately?
- Were other variables suitably controlled (i.e. constant room temperature for all trials, or same quantity of variables used)?
- Were any systematic errors (such as zero error and technical defaults) suitably minimised and/or controlled?
Accuracy:
Accuracy is the closeness of a measured value to its true values and allows valid conclusions to be made. The accuracy of data is considered by analysing the precision of measurement and errors. One way to enhance the accuracy of data is to use measuring instruments with relatively a relatively small limit of reading. Errors in experimental design harm the accuracy of data and ultimately impede both the reliability and validity of the investigation. Several causes of inaccurate results are:
- The use of measuring instruments with an unsuitable limit of reading.
- The use of measuring instruments with errors in their function (may be due to old age or is not zeroed before use creating a systematic error).
- Random errors, such as human reflexes/reaction or parallax error (e.g. reading the volume of liquid in beaker from above the beaker)
- The quantity that is being measured may be constantly changing or has non-uniformity (e.g. length of head-hair)
Random Errors:
Random errors are caused by unknown and unpredictable changes in the experiment (e.g. due to the instruments or environmental conditions). These are statistical fluctuations in both directions about the true value and thus repetition and statistical analysis can reduce such effects. Random errors can be reduced in several ways:
- Make sure you know how to read the scales on the instruments, and that you align yourself correctly each time you take a measurement.
- Take multiple measurements (repetition increases reliability), before calculating an average for the gathered quantitative data (reducing random error).
Systematic Errors:
Systematic errors are caused by measuring instruments either being used incorrectly or possessing damage to their structure and thus harming their function. Systematic errors limit accuracy and thus the following steps should be taken to reduce such errors:
- Check all instruments against a standard before conducting the investigation. Zero settings should be checked and adjusted (i.e. calibrated).
- Instructions for the use of instruments should always be available, read and followed carefully.
- Corrections for instrument bia should be made, if necessary.