1.9.2 Accuracy Accuracy is a relatively straightforward concept and refers to the absence of errors in the collecting, transcribing, coding and analysis of data. It is about taking care when doing the research and avoiding mistakes.
When collecting your own data it is not unusual to make mistakes but checking can reveal inaccuracies and minimise the impact on the research. When using secondary data, although the researcher can take care in subsequent reanalysis, there is usually no way of knowing if the original data collection was error free.
Accuracy is also an issue whe it comes to retrospective measurement. Notoriously, people asked to estimate their alcohol consumption over the past week overwhelmingly underestimate actual consumption. Similarly, asking, 'How much television did you watch last night?' would require retrospective estimates and would not be as potentially accurate as, for example, someone with an electronic handset designed to record how much and which channels are watched: the user clicking on or off whenever they start and stop watching. This is potentially highly accurate and precise (see below) but one should not rule out the possibility that the user may forget to click on or off and thus provide erroneous readings.
One confusion arises when 'reliability' is also defined as 'accurate measurement'. While an accurate measurement is necessary when implementing a reliable test or measurement instrument, it is not the accuracy of the measurement process that determines the reliability of the instrument, it is the design of the instrument itself to which reliability refers. A ruler is a reliable measure of length and it does not stop being reliable if someone uses it without taking care so that the measures are inaccurate.
It is important to distinguish between accurate and unbiased (See Section 126.96.36.199) . A researcher may make mistakes in the data collection and analysis process and these may or may not have a significant effect on the results. In a sense, these are arbitrary and could, through rigorous checking, be reversed.
Bias, however, in social research is usually restricted to a systematic distortion that is independent of inaccurate working. Bias, which is discussed in more detail in Section 1.10.3, refers to an approach to data collection that excludes some part of the population that should have been included.
For example, a sample supposedly of the general population drawn from names in the telephone directory excludes: all those people who have opted to be ex-directory; do not have a land-line and therefore are not listed; or there are other people in the household of the person who is listed in the directory, which more often than not tends to be the senior male in the family. Bias is, thus, about the representativeness of the data.
One should also distinguish between accurate and precise. Accuracy is about not making mistakes. Precision is about measuring something to a specified degree. In describing a crowd situation is it necessary to know that the crowd was about 200 people or is it important to know that it consisted of 192 people. Is it important to know that the crowd was about 200 people, mostly young white males, or is it necessary to identify the gender, age and ethnicity of all 192 people who made up the crowd? The degree of precision is a function of the purpose of the study and one should avoid ever-more fine-grained precision for the sake of it.