M

Statistical Research Vocabulary and Study Design Notes

Observational vs Experimental Research

  • Primary purpose of statistical research: answer a research question (examples provided):
    • Does aspirin reduce the risk of heart attacks?
    • Is one brand of fertilizer more effective at growing roses than another?
    • Is fatigue as dangerous to a driver as the influence of alcohol?
  • Involves investigation of the relationship between at least two variables.
  • Proper study design ensures the resulting data is reliable and accurate.

Observational Research (Non-experimental)

  • The researcher merely observes what is currently happening or what has happened in the past (records) and tries to draw conclusions based on these observations.
  • Little or no interaction with subjects.
  • No manipulation is performed – do not try to change anything about subjects.
  • No cause and effect relationship can be determined.
  • Typically occurs in natural settings.
  • Data collected is often qualitative, but can be quantitative or both.
  • Can be expensive and time consuming.
  • Prevalent in Social Sciences and Marketing (Business).
  • Examples:
    • Surveys (telephone, mailed questionnaire, personal interview).
    • Case studies.

Experimental Research (aka Experiment)

  • The researcher manipulates one of the variables and tries to determine how the manipulation influences other variables.
  • Interaction with subject usually occurs; modifications on subject occur.
  • May occur in unnatural settings (labs or classrooms).
  • Can identify a cause-and-effect relationship.
  • Used in sciences such as Sociology, Psychology, Biology, Chemistry, Physics, and Medicine.
  • Examples:
    • Clinical trials of new medications or treatments.
    • Lab experiments.

The Variables

  • Independent Variable (aka Explanatory Variable)
    • Is the variable or thing that is being manipulated or changed by the researcher.
    • In a sense, this variable EXPLAINS what is happening in the experiment.
    • Often represented by the letter x.
  • Dependent Variable (aka Response Variable)
    • Is the result (outcome) of a treatment or manipulation of the independent variable.
    • Variable about which the researcher is asking a question.
    • Often represented by the letter y.

The Setup

  • Experimental Units
    • Single object or individual to be measured.
    • Also called Subjects.
  • Treatments
    • Different values of the explanatory variables.
    • Typically have at least 2 treatments.
  • Control Group
    • Group of subjects that do NOT receive any treatment or receive a PLACEBO (neutral or fake treatment).
    • Provides a baseline for comparison.
  • Treatment Group
    • Group of subjects that receive a treatment.
    • May have multiple treatment groups (one for each treatment).
  • Random Assignment
    • Subjects are randomly assigned to a group prior to the start of the experimental research study.
  • Blinding
    • Subjects do not know to which group they are assigned (BLINDING) to prevent or limit the effects of “suggestion.”
    • Researcher and subjects do not know which group is which (DOUBLE BLINDING) to prevent or limit bias in the research.

Caution!!! Lurking Variables

  • Lurking variables are variables not included as part of the research, but influence or affect the interpretation of the relationship.
  • May be unknown variables and thus not controlled.

Examples (Classification Practice from the Transcript)

  • Scenario items to classify observational vs experimental:
    • Subjects randomly assigned to two groups; one group given an herb and the other a placebo; after a year, number of respiratory tract infections compared.
    • Classification: Experimental (random assignment; placebo control).
    • GHC President wants to know the average age of students enrolling in Fall 2020; uses SCORE student records to determine the age of each student.
    • Classification: Observational (uses existing records; no manipulation).
    • A researcher stands at the entrance to see if the color of the automobile a person drives is related to running red lights.
    • Classification: Observational (no manipulation; seeking association).
    • A dentist wants to research the effects of a new dental material used to make fillings.
    • Classification: Experimental (testing a new material).

Scenario 1 (Aspirin and Heart Attacks) – Page 13 Transcript

  • Study design: Four hundred men aged 50–84 recruited and randomly divided into two groups; one group takes aspirin, the other a placebo; one pill per day for three years; final count of men who had heart attacks.
  • Identify the variables:
    • Explanatory (independent) variable: ext{Treatment assigned} = ext{aspirin vs placebo}
    • Response (dependent) variable: ext{occurrence/count of heart attacks in each group}
    • Treatments: ext{Aspirin}, ext{Placebo}
  • Blinding: The men do not know which pill they take, so this is a blinded study for participants; text specifies blinding of participants, which corresponds to single-blind.
    • Note: The transcript does not specify whether researchers are blinded; thus it is not necessarily double-blind based on the given information.

Scenario 2 (Smell and Learning) – Page 14 Transcript

  • Study design: Subjects complete mazes multiple times while wearing masks; mazes repeated three times with floral-scented masks and three times with unscented masks; assignment to wear the floral mask first or last three trials is random; recorded metrics include maze completion time and the subject’s impression of the mask’s scent (positive, negative, neutral).
  • Identify the variables:
    • Explanatory variable: ext{Mask scent condition} = ext{floral vs unscented}
    • Response variables:
    • Primary: ext{Time to complete the maze (in time units)}
    • Secondary: ext{Impression of scent} ext{ ∈ \{positive, negative, neutral\}}
    • Treatments: The two scent conditions (floral vs unscented) applied across trials; the random assignment of floral mask to the first three vs last three trials is a design choice to control for order/fatigue effects.

Key Takeaways for Study Design

  • Observational studies cannot establish causation due to potential lurking variables and lack of controlled manipulation.
  • Experimental studies actively manipulate an independent variable to observe effects on a dependent variable, enabling causal inferences when well-controlled.
  • Random assignment helps ensure groups are comparable and reduces selection bias.
  • Blinding minimizes bias:
    • Blinding (single) = participants are unaware of group assignment.
    • Double blinding = both participants and researchers are unaware of group assignment.
  • Control groups provide a baseline to compare against treatment groups.
  • Lurking variables can confound results if not identified and controlled for; they are variables not included in the study design but that influence the outcome.
  • In practice, many real-world studies blend observational and experimental elements, require careful interpretation, and consider ethical, philosophical, and practical implications (e.g., feasibility, consent, risk considerations).

Formulas and Notation Recap

  • Independent variable (explanatory): X
  • Dependent variable (response): Y
  • Treatments are values or levels of the independent variable (e.g., X = ext{Aspirin}, X = ext{Placebo}).
  • For a given study, the relationship of interest can be summarized in data such as counts, means, probabilities, etc., often requiring statistical inference to determine if observed differences are statistically significant.