Measurement methods in Research
As with the selection of IVs, the selection of dependent variables is often complicated by practical constraints. For example, if we are investigating the impact of alcohol consumption on road fatalities, we may manipulate the independent variable straightforwardly (by getting experimental groups to consume different quantities of alcohol). But it would be irresponsible (and illegal) to then get the participants to drive down a busy street so that we can count how many pedestrians they knock down! To get round this, we may ask the high alcohol group to consume only a few beverages. But there are two problems with this. First, alcohol may only affect driving behaviour when more than a few beverages are consumed. Second, our dependent variable (number of pedestrians killed) will not be sufficiently sensitive to detect the independent variable’s impact. In other words, we may have good reason to think that alcohol could impair driving performance, but the degree of impairment may not (fortunately!) be so profound as to cause a detectable increase in the number of deaths caused.
To deal with this, we therefore have to select dependent variables that are both relevant to the outcome we have in mind and sensitive to the independent variable. In the case of drink-driving, we may look at participants’ reaction time, because we believe that this is a critical determinant in driving safety and is likely to be a sensitive enough variable to detect an impairment in driving performance due to alcohol. We can then design and carry out a study in the laboratory, measuring the impact of alcohol consumption on reaction time. In our attributional style example, too, it is unlikely that our manipulation of the independent variable will have a dramatic impact on the participants’ depression. So if our dependent variable was the number of participants who need to be treated by a clinical psychologist, our experiment is very unlikely to uncover any effects. To get around this problem, we could administer a depression inventory, in which we ask the participants a battery of questions (e.g. ‘Are you self-confident?’, ‘Do you feel hopeless about the future?’) in order to measure their susceptibility to depression. We could then test our hypothesis by seeing whether scores on the depression inventory revealed a higher susceptibility to depression among participants who had been encouraged to make internal attributions.
The psychologist S.S. Stevens developed a famous distinction between forms of data that psychologists can deal with. The four types he came up with are nominal, ordinal, interval and ratio measures.
[Stanley Smith Stevens (1906–73) made significant contributions to several areas of psychology. He was an expert on the psychophysics of hearing and was interested in measurement and experimental psychology. Stevens set out to redefine psychological measurement by changing the perspective from that of inventing operations (the physical view) to that of classifying scales (a mathematical view). He also discovered that methods such as ‘just noticeable differences’, rating scale categories and paired comparisons produce only ordinal scales. Stevens’ most oustanding contribution was his successful argument that there are different kinds of scales of measurement, being the first to define and discuss nominal, ordinal, interval and ratio scales.]
Nominal measures
The data collected in this way are in the form of names, which can be categorized but cannot be compared numerically in any way. Examples include genders, countries and personality types.
Ordinal measures
These can be ranked in some meaningful way. Examples are the placings obtained by competitors in a race or an ordered set of categories (e.g. low stress, moderate stress and high stress).
Interval measures
Numerical measures without a true zero point are called interval measures, and cannot be used to form ratios. An example is temperature. The zero point has been arbitrarily chosen to be the freezing point of water rather than absolute zero (where there is no temperature), and it is simply not true that 40 degrees Celsius is twice as hot as 20 degrees Celsius. Similarly, it would not make sense to say that someone who responded with a ‘6’ on the attribution scale above was twice as much of an externalizer as someone who responded with a ‘3’.
Ratio measures
Full numerical measures with a true zero point are ratio measures. Psychologists frequently assume that scores obtained from psychological measurement can be treated as ratio measures. But this assumption is not always justified.
As with the selection of IVs, the selection of dependent variables is often complicated by practical constraints. For example, if we are investigating the impact of alcohol consumption on road fatalities, we may manipulate the independent variable straightforwardly (by getting experimental groups to consume different quantities of alcohol). But it would be irresponsible (and illegal) to then get the participants to drive down a busy street so that we can count how many pedestrians they knock down! To get round this, we may ask the high alcohol group to consume only a few beverages. But there are two problems with this. First, alcohol may only affect driving behaviour when more than a few beverages are consumed. Second, our dependent variable (number of pedestrians killed) will not be sufficiently sensitive to detect the independent variable’s impact. In other words, we may have good reason to think that alcohol could impair driving performance, but the degree of impairment may not (fortunately!) be so profound as to cause a detectable increase in the number of deaths caused.
To deal with this, we therefore have to select dependent variables that are both relevant to the outcome we have in mind and sensitive to the independent variable. In the case of drink-driving, we may look at participants’ reaction time, because we believe that this is a critical determinant in driving safety and is likely to be a sensitive enough variable to detect an impairment in driving performance due to alcohol. We can then design and carry out a study in the laboratory, measuring the impact of alcohol consumption on reaction time. In our attributional style example, too, it is unlikely that our manipulation of the independent variable will have a dramatic impact on the participants’ depression. So if our dependent variable was the number of participants who need to be treated by a clinical psychologist, our experiment is very unlikely to uncover any effects. To get around this problem, we could administer a depression inventory, in which we ask the participants a battery of questions (e.g. ‘Are you self-confident?’, ‘Do you feel hopeless about the future?’) in order to measure their susceptibility to depression. We could then test our hypothesis by seeing whether scores on the depression inventory revealed a higher susceptibility to depression among participants who had been encouraged to make internal attributions.
The psychologist S.S. Stevens developed a famous distinction between forms of data that psychologists can deal with. The four types he came up with are nominal, ordinal, interval and ratio measures.
[Stanley Smith Stevens (1906–73) made significant contributions to several areas of psychology. He was an expert on the psychophysics of hearing and was interested in measurement and experimental psychology. Stevens set out to redefine psychological measurement by changing the perspective from that of inventing operations (the physical view) to that of classifying scales (a mathematical view). He also discovered that methods such as ‘just noticeable differences’, rating scale categories and paired comparisons produce only ordinal scales. Stevens’ most oustanding contribution was his successful argument that there are different kinds of scales of measurement, being the first to define and discuss nominal, ordinal, interval and ratio scales.]
Nominal measures
The data collected in this way are in the form of names, which can be categorized but cannot be compared numerically in any way. Examples include genders, countries and personality types.
Ordinal measures
These can be ranked in some meaningful way. Examples are the placings obtained by competitors in a race or an ordered set of categories (e.g. low stress, moderate stress and high stress).
Interval measures
Numerical measures without a true zero point are called interval measures, and cannot be used to form ratios. An example is temperature. The zero point has been arbitrarily chosen to be the freezing point of water rather than absolute zero (where there is no temperature), and it is simply not true that 40 degrees Celsius is twice as hot as 20 degrees Celsius. Similarly, it would not make sense to say that someone who responded with a ‘6’ on the attribution scale above was twice as much of an externalizer as someone who responded with a ‘3’.
Ratio measures
Full numerical measures with a true zero point are ratio measures. Psychologists frequently assume that scores obtained from psychological measurement can be treated as ratio measures. But this assumption is not always justified.
No comments:
Post a Comment