Lec. 3 Beginning Research

Question 1

Beginning Research

Answer 1

THE RESEARCH CHALLENGE – Psychologists need to measure abstract ideas like “truth” (Forensic Psychologist) or “Intelligence” (cognitive psychologists), but they can only observe behaviors that they think are related to those things. So how would you go about measuring truth, intelligence, depression, or any other hypothetical construct?

1. Start with an OPERATIONAL DEFINITION – Which describes an abstract concept in terms of how we measure it.
   
   • Ex. a clinical psychologist can’t just look at an individual and know the degree of depression the person is experiencing.
   • He must use some diagnostic measurement tool – like a test.
     
     • The score on a test would be our operational definition.
     • A high score reflects greater depression and a low score reflects less. This is how the Beck Depression Inventory ( BDI), the most common ‘Depression Test’ is used.
   • Having an operational definition allows us to test our HYPOTHESIS – a testable prediction involving two or more variables where an INDEPENDENT VARIABLE is HYPOTHESIZED to affect a DEPENDENT VARIABLE.
     
     • INDEPENDENT VARIABLE – the variable controlled by the researcher in an experiment.
       
       • MUST have at least TWO levels to compare.
         
         • Ex. EXPERIMENTAL GROUP vs. CONTROL GROUP
     • DEPENDENT VARIABLE – the variable being affected by the INDEPENDENT VARIABLE.
     • WITHIN SUBJECT EXPERIEMENT – different levels of the independent variable are applied to the same subject.
       
       • A group receives 5 mg of the drug per day for 6 months. The same group then receives 10 mg of another drug in a different 6 month period.
     • BETWEEN GROUP EXPERIMENT – different levels of the independent variable are applied to different groups.
       
       • Group A receives 5 mg of the drug per day for 6 months. Group B receives 10 mg of another drug over the same 6-month period.

Question 2

Sampling

Answer 2

PROBABILITY SAMPLING – everybody in your interested population has an EQUAL chance of participating in your study.

• Rarely used because it would be prohibitively expensive to include people from every country, of all ages, with diverse backgrounds.
• Instead, researchers settle for samples of college students mostly, because they are the ones available (Availability Bias).
• The CRITICAL QUESTION is whether that difference is important in your research.
  
  • Ex: If you are studying reaction times to the appearance of a visual stimulus on a screen, college students may be very similar to other people.
  • HOWEVER, if you want to study voting preferences, college students may be dissimilar to people not in college. In our research, we often do not know how well our sample mirrors the population of interest.

NON-PROBABILITY SAMPLING

• Most psychological research does not involve probability sampling, which means we often do NOT know to whom our research results generalize – other than other college students (LOW EXTERNAL VALIDITY)
• NON-SAMPLING ERROR – occurs when people who should be included in a sample are not.
  
  • This is the biggest problem with nonprobability samples because nonprobability samples are not representative of the population as a whole.
  • Experimental psychology involves students almost exclusively because they are easy to obtain and pleasant to work with.

CONVENIENCE SAMPLING (BIAS) – when researchers rely on a population because it is easy or available.

• Tend NOT to reflect the desired population.
• That said, comparing experimental groups, we don’t necessarily care about the exact measurement each group gets, but rather patterns of differences.
  
  • i.e. How a change in the INDEPENDENT VARIABLE affected one group RELATIVE to the other.

Question 3

Reliability, Validity

Answer 3

RELIABILITY – If a measurement is RELIABLE, then repeated measurements on the same person should result in similar outcomes each time.

• Reliability relates to consistency and repeatability in the results.
• Just because an intelligence test is reliable, we shouldn’t automatically assume that it shows VALIDITY – that the test is useful.
  
  • Tests can be reliable without giving us any useful information.
  • Ex. We could use your height as an estimate of your intelligence; the measurement would be reliable, but it wouldn’t be valid.

TEST-RETEST RELIABILITY – When a measurement yields similar results with repeated testing.

SPLIT-HALF RELIABILITY – when questions from subcomponents of a test lead to similar results.

INTERRATER RELIABILITY – When data from two different observers tend to agree.

VALIDITY – Validity relates to the question of whether measurements provide information on what we really want to measure.

• In order for a measurement to be VALID, it must ALSO be RELIABLE.
  
  • But a RELIABLE measurement might NOT necessarily be VALID.
• Another way of saying research results are MEANINGFUL is to say that they are VALID.
• VALIDITY goes from low to high on a spectrum.
• VALIDITY may legitimately and predictably change in different situations.
• A given experiment may show high levels of some types of VALIDITY, but lower levels of different types.
• TYPES OF VALIDITY:
  
  • INTERNAL VALIDITY – Only the INDEPENDENT VARIABLE has an influence on the DEPENDENT VARIABLE.
    
    • Extraneous and confounding variables have been eliminated from your methodology.
    • When research shows internal validity, it means that you:
      
      • Started your project with groups that were comparable with respect to the DEPENDENT VARIABLE ( or equated them using statistical procedures).
      • Eliminated confounds and extraneous variables, manipulated an INDEPENDENT VARIABLE that possessed CONSTRUCT VALIDITY – refers to whether a scale or test measures the construct adequately.
      • Held everything but the INTERNALLY VALID treatments constant across your groups
      • Measured your participants’ behaviors accurately on another variable that had CONSTRUCT VALIDITY.
• EXTERNAL VALIDITY – the ability to generalize your findings to a broader population and situation.

Question 4

Random Selection and Random Assignment

Answer 4

RANDOM SELECTION – refers to how the sample is drawn from the population as a whole where everybody in the population has a specific and equal probability of being included in the research.

• Used in some research, mostly surveys, otherwise, college students are typically used.
• Increases the representativeness of the sample.

RANDOM ASSIGNMENT – refers to how the participants are assigned to either the experimental or control groups.

• Uses an objective and unbiased (i.e. random) selection strategy so that any individual has an equal chance of ending up in either group.
• This creates similar groups for comparison.
• But Random Assignment is not perfect and is subject to RANDOM SAMPLING ERROR – when, by total chance, random assignment results in a biased sample.

Question 5

Scales of Measurement

Answer 5

SCALES OF MEASUREMENT

• NOMINAL SCALE – descriptive, simple categorization.
  
  • Ex: 80% Male or 20% Female, 52% Democrat or 48% Republican
  • Simplest form of measurement.
  • Nominal scales are mutually exclusive and collectively exhaustive.
  • There isn’t much arithmetic we can do with these numbers other than counting. We can differentiate among the categories themselves only descriptively.
• ORDINAL SCALE – There is order (one is larger than the other), but no indication of space between the items.
  
  • Ex: A typical example of measurements on an ordinal scale are RANKS. Somebody ranks 1st, someone 2nd, someone 3rd, but there is no indication about how far apart they are.
• INTERVAL SCALE – There is order AND there is EQUAL DISTANCE between values.
  
  • Ex: Age (1,2,3,4…) or IQ or height – has NO ZERO value
• RATIO SCALE – There is order AND there is EQUAL DISTANCE and also has a zero value.
  
  • Ex: minutes of music.