Non-Parametric Evaluation and Qualitative Research Flashcards

Non-Parametric Evaluation and Qualitative Research

Non-Parametric Statistics

• Require fewer assumptions and conditions compared to parametric statistics (t-tests, ANOVAs).
• Used when data violates parametric assumptions, most notably when data is not normally distributed.
• Considered less powerful than parametric tests.

When to Use Non-Parametric Tests:

• Non-normal distribution.
• Use a normality test to determine (Kolmogorov-Smirnov, Anderson-Darling, and Shapiro-Wilk).
• Outcomes are ordinal or rank ordered.
• Presence of outliers.
• Limited levels of detected differences.

Variables

Nominal

• No inherent relationship between categories.
• Examples: Yes/No, Male/Female.

Ordinal

• Rank order without necessarily equal distance between points.
• Example: 1st place -> 2nd place -> 3rd place.

Interval

• Numerically equal distance between points.
• Example: 95 is a better grade than 78.
• Associated with parametric data.

Ratio

• Interval value with a baseline.
• Zero actually means zero or absence of the variable.

Levels of Measurement

• Nominal: Named (e.g., eye color).
• Ordinal: Named, natural order (e.g., level of satisfaction).
• Interval: Named, natural order, equal interval between variables (e.g., temperature).
• Ratio: Named, natural order, equal interval between variables, has a "true zero" value (e.g., height).

Independent Samples

• Chi-Square test
• Mann-Whitney U Test: Used similarly to the parametric independent/unpaired samples t-test.

Chi-Square Test (\chi^2)

• Difference between distribution of 1 sample compared to a hypothetical or known distribution.
• Frequencies represent individual counts/events.
• Categories are mutually exclusive (e.g., tired: yes/no).
• The larger the difference between observed and expected, the larger the \chi^2 value.
• Formula: \chi^2 = \Sigma \frac{(Observed - Expected)^2}{Expected}
• Compare \chi^2 to tabled data (power level: \alpha = 0.05).
• Value must equal or exceed tabled value to reject the Null Hypothesis (H_0).
• H_0 states that no differences exist between the observed and expected counts.
• Groups must be mutually exclusive.

Chi-Square Example: Coin Flip

• Expected: 50 heads, 50 tails out of 100 flips.
• Observed: 44 heads, 56 tails.
• Calculation:
  • \chi^2 = \frac{(44-50)^2}{50} + \frac{(56-50)^2}{50}
  • \chi^2 = \frac{36}{50} + \frac{36}{50}
  • \chi^2 = 0.72 + 0.72 = 1.44
• Degrees of freedom (df) = (# groups – 1) = 1.
• Tabled critical value for \alpha = 0.05 and df = 1 is 3.84.
• Since 1.44 < 3.84, we fail to reject the null hypothesis.

Minimum Head/Tail Distribution to Exceed Critical Value

• 40 Heads:60 Tails
• \chi^2 value = 4.0.

Chi-Square Example: Therapy Settings

• Categories: Outpatient, Acute, Home Health, School System.
• Data:
  • Last 10 years (Expected): 26, 12, 5, 2 (Total 45)
  • Class of 2019 (Observed): 21, 16, 7, 1 (Total 45)
• \chi^2 = 3.59
• Critical value for \alpha = 0.05, df = 3 is 7.81.
• The Class of 2019 is not significantly different from the last 10 years.

Correlated Samples

• Pairs or Pre/Post Data
• Accounts for direction of difference & magnitude of difference between samples

Wilcoxon Matched Pairs Signed Rank Test

• Accounts only for direction of difference between samples.
• Less powerful than Wilcoxon Sign Test.

Critical Values of the Wilcoxon Signed Ranks Test

• Table provided with critical values for different n values and alpha levels (one-tailed and two-tailed tests).

Wilcoxon Matched Pairs Signed Rank Test Example

• Knee strength MMT from 0-12 in 2 different positions.
• Absolute value of differences ranked; shared ranks are averaged.
• Use the combined ranks of the lesser number of signs.
• If T = -1 and the critical value from the table = 2 then Because our test does not exceed this value it IS SIGNIFICANT. This is opposite from how we typically use critical values.

Other Non-Parametric Tests

• Friedman 2-way ANOVA: > 2 related samples
• Kruskal-Wallis: 3 or more unrelated samples
• Spearman’s Rank Order Correlation Coefficient: Degree of relationship between 2 ordinal variables.

Spearman’s Rank Order Correlation Coefficient Example

• Ranks of students in Math and English.

Qualitative Research

Qualitative Vs Quantitative

Types of Qualitative Research

Case Report

• Complete perspective of an individual.
• Unusual or unique findings OR unique therapeutic intervention.
• Very detailed with rich descriptions.
• Very individualized.
• You will write up at least 1 of these before your time here is done based on your clinical experiences!

Narrative/Life History

• Chronological sequence of person’s perceptions & experiences.
• Self-disclosures or biographical.
• Life span & multiple events.

What can these tell us?

• Responses to disability.
• Behavior around chronic conditions.
• Intervention timing.
• Availability of services.
• Support systems.
• Coping skills.

Ethnography

• Typically sociology & anthropology: systematic study of individual cultures.
• Events, situations, and interactions.
• Long term association of the researcher with persons or a group.
• Participates but is distinctly separate.
  • differences in services due to setting
  • interpersonal interactions
  • patient compliance
  • environmental factors
  • professional socialization
  • decision-making process
• Examples: Studies on wild chimpanzees, coming of age in Samoa, physiotherapists' perceptions of their interactions with patients on a chronic pain unit.

Grounded Theory

• Researcher collects, codes, and analyzes data simultaneously.
• Theory is “grounded” in the observations, not a preconceived hypothesis.
• Data undergoes constant comparative analysis and is always developing → “Refined”.

Applications of Grounded Theory

• Influence of culture.
• Development & nature of interpersonal relationships.
• Efficacy of quality assurance mechanisms.
• Identification & handling of ethical issues.
• Influence of technology.
• Development of therapeutic interventions.

When to Use Qualitative Research Methods?

• Not appropriate when:
  • Numerical data is needed for statistical analysis.
  • Testing or verifying hypotheses.
  • When confounding variables cannot be eliminated for logistical or ethical reasons.

Getting Started

• Identify a topic
  • Experience
  • Literature
• Posing a question
  • Global
  • Specific - setting, persons, events, or behaviors
• Review the literature
  • Constantly evolving
  • “Has this been asked or answered already?”

Collecting Data

• Observations
  • Direct
  • Indirect
  • Unobtrusive data?
• Interviews
  • Extended period of time to complete

Observations

• Training is required.
• Direct Observation
  • Goal to be inconspicuous OR embed in group with “participant observer”.
• Settings
• Informed Consent?

What to Observe?

• Physical Environment
• Social Environment
• Events
• Amount of Observations
  • as much as it takes, varied
• Recording Observation

Interviews

• Types of Interviews
  • Unstructured, Semi-structured, Structured
• Types of Questions
  • Behavior
  • Opinion
  • Feeling
  • Knowledge
  • Sensory
  • Background
  • *avoid leading questions
• Conducting Interview
  • Active listening, recording

Unobtrusive Data

• Physical Traces
  • Erosion, deposits
• Written Materials
  • Records, protocols, etc.
• Observation
  • Public, hidden → Ethical gray area?
• Recordings
  • Audio, video, photographs
  • “Disguised observation”