Objective Measurement of Subjective Phenomena

4. Nature of the Construct

Exercise 1

In this exercise the user should select the most appropriate dimension for each research scenario.

Instructions:
For each of the following research scenarios, identify the nature of the personal characteristic to be measured.  Select the most appropriate dimension category below the scenario.

Dimensions:
Dimension 1

  • Continuous
  • Dichotomous
  • Polytomous

Dimension 2

  • Broad
  • Narrow

Dimension 3

  • Context independent
  • Context dependent

Dimension 4

  • Trait constructs
  • State constructs

Dimension 5

  • Acute
  • Chronic

Dimension 6

  • Growth
  • Decline

Research Scenarios:
Scenario A:  A researcher wants to see if a new treatment is effective in managing chronic pain. 
On which dimensions should the researcher concentrate when constructing a set of items?
Scenario B:  A physician is interested in developing an instrument to distinguish between general manic (e.g., excited, agitated) behavior and manic behavior that seems to occur only when a person is facing impending medical procedures.
Which dimensions seem most relevant when considering ways to measure these separable constructs?
Scenario C:  A research team is beginning a study of subjects in middle adulthood (e.g. around age 45 years), intending to follow them longitudinally for 30 years to track changes in their personality, everyday functioning, and cognitive abilities.
Which two dimensions are most relevant when designing assessment instruments for this study?
Scenario D:  In a multisite intervention study, the researchers have developed a tiered intervention, in which all families receive vitamin supplements, some families also receive home visits by a nurse, and a subset of these latter families also have day care provided for their young infants beginning at age 6 months.  The 5-year intervention is designed to lead to higher levels of positive social and peer behaviors by children both at home and in school during the early elementary grades (grades 1-3).
Which dimensions seem most important to stress in assessments?
Scenario E:  An investigator wants to distinguish, in a very general way, between different levels of maltreatment experienced by children.  In this research project, distinguishing among three levels of maltreatment by parents (e.g., none, some, a great deal) seems the most relevant.
On which dimension or dimensions should the investigator concentrate?
Scenario F:  A research team has been assigned the task of differentiating general anxiety and test anxiety.  The team will assess college students on these constructs each week during a semester of study, where weeks 5, 10 and 15 correspond to the weeks that students will have tests in their courses.  Which dimensions seem most relevant to consider?

Answers:
Scenario A:  A researcher wants to see if a new treatment is effective in managing chronic pain. 
On which dimensions should the researcher concentrate when constructing a set of items?
Dimension 1:  A continuous rating line or a rating scale with many points (e.g., a rating on a scale ranging from 0 = no pain to 10 = excruciating pain) would be more helpful than a simple dichotomy (e.g., no pain vs. some pain) or polytomous scale (e.g., no, some, very high), so that level of pain could be well differentiated.  A larger number of rating scale points would allow patients to register more subtle changes in their levels of chronic pain.
Dimension 5:  The pain suffered by persons to be assessed is chronic, so the items should emphasize the level of pain suffered over an appropriate time frame (e.g., since the last doctor visit, during the last week, during the last month), rather than the level of pain experienced right at the present time.

Scenario B:  A physician is interested in developing an instrument to distinguish between general manic (e.g., excited, agitated) behavior and manic behavior that seems to occur only when a person is facing impending medical procedures.
Which dimensions seem most relevant when considering ways to measure these separable constructs?
Dimension 3:  Perhaps the most obvious need is to consider the context within which the manic behaviors are shown.  Certain items should assess manic behaviors as seen in a person’s general life circumstances (e.g., during the past two weeks), and other items should assess relevant behaviors in the minutes or hours preceding medical procedures the person fears.  Some persons may have relatively low levels of manic behaviors in general, but show greatly increased levels of such behaviors in particular situations.
Dimension 4:  Manic behaviors tend to fluctuate some over time – showing some state-like properties – so the general manic behaviors should be assessed as they change over time.  But, the specific context of the medical procedures should be incorporated into the scale designed to assess the exacerbating effects of the one context assumed to heighten the behaviors by some individuals.

Scenario C:  A research team is beginning a study of subjects in middle adulthood (e.g. around age 45 years), intending to follow them longitudinally for 30 years to track changes in their personality, everyday functioning, and cognitive abilities.
Which two dimensions are most relevant when designing assessment instruments for this study?
Dimension 4:  Personality and cognitive abilities often are assumed to be trait constructs, which change relatively slowly over time.  On the other hand, everyday functioning may display some variability across time, with persons having “good days” versus “bad days” that can be distinguished by some state-like changes in quality of functioning from day to day.
Dimension 6:  In any longitudinal study, it is important to develop measures that allow for detection of growth (or improvement) and decline.  Thus, measures should have sufficient ceiling (so individuals can exhibit improvements in performance) and sufficient floor (so that the extent of declines can be identified).

Scenario D:  In a multisite intervention study, the researchers have developed a tiered intervention, in which all families receive vitamin supplements, some families also receive home visits by a nurse, and a subset of these latter families also have day care provided for their young infants beginning at age 6 months.  The 5-year intervention is designed to lead to higher levels of positive social and peer behaviors by children both at home and in school during the early elementary grades (grades 1-3).
Which dimensions seem most important to stress in assessments?
Dimension 2:  The social and peer domains are quite broad, with many distinctions possible among dimensions to assess.  The research team should try to ensure that all relevant aspects of each domain are assessed, so that effects are seen in general improvements, rather than in extremely narrow subdomains with little generality.
Dimension 3:  The intervention is designed to be effective across the school and home environments (i.e., not to be context dependent), so it will be important to assess relevant behaviors in both environments to demonstrate that effects are not seen only in one of the two environments.
Dimension 6:  The issue of adequate ceiling in measurements so that growth can be charted is also of importance.  The age range from 6 months to around 8 or 9 years (3rd grade) is one during which tremendous changes in the forms of behavior are seen.  As a result, great care must be taken to ensure that measures with the same name assess the same construct at different ages and that enough “room to grow” on the assessment scales has been provided.

Scenario E:  An investigator wants to distinguish, in a very general way, between different levels of maltreatment experienced by children.  In this research project, distinguishing among three levels of maltreatment by parents (e.g., none, some, a great deal) seems the most relevant.
On which dimension or dimensions should the investigator concentrate?
Dimension 1:  A polytomous measure of maltreatment by parents – assessed at levels corresponding to no, some, and a great deal of maltreatment – may be sufficient to index the negative experiences by a child.  This would enable the researcher to contrast groups on the basis of having no maltreatment versus maltreatment and then, for those who experienced maltreatment, whether some versus a great deal of maltreatment had notable differential effects.

Scenario F:  A research team has been assigned the task of differentiating general anxiety and test anxiety.  The team will assess college students on these constructs each week during a semester of study, where weeks 5, 10 and 15 correspond to the weeks that students will have tests in their courses.  Which dimensions seem most relevant to consider?
Dimension 2:  The general anxiety construct is a more general, broad construct, whereas the notion of text anxiety is a very narrow one.  Monitoring trends in these two constructs, which vary considerable in breadth, would allow the researchers to see whether text anxiety rises and falls with general anxiety or whether the two fluctuate independently over time.
Dimension 5:  The general anxiety construct represents a more chronic state – some individuals are more anxious than others, and these general levels tend to remain relatively stable over time.  In contrast, test anxiety is likely to spike in the weeks during which students must take a large number of tests, and these acute spikes would be shown by the subset of students who are prone to this form of anxiety.