EBP Final: Inference

Parametric Tests

Uses interval/ratio data and normally distribution population

Non-Parametric tests

Uses nominal/ordinal data and data is not normally distributed

What are the parametric tests of relationships?

Pearson's correlation, ICC, Linear regression

Pearson's Correlation

Relationship between two variables

Example: Investigating the relationship between shoulder internal rotation ROM and overhead throwing velocity in pitchers.

ICC

Measures inter-rater reliability of repeated measures

Example: You want to know if two PTs measure hamstring flexibility the same way using a goniometer.

ICC = 0.92 → Excellent reliability

Linear Regression

Predicts the value of one variable (dependent) based on one or more others (independent). Cause and effect

Simple (one predictor)

Multiple (Two or more predictors)

Example: Predicting whether a stroke patient will be discharged home vs to SNF based on FIM score and gait speed.

What are the non-parametric tests of relationships?

Spearman Rank, Kappa, Logistic Regression

Spearman Rank Correlation

Opposite of Pearson, measures the relationship of two variables

Example: Ranking pain levels on a 0–10 scale vs. patient satisfaction ranked from 1–5.

Kappa

Opposite of ICC, measures inter-rater reliability of repeated measures

Example: Two clinicians grade shoulder abduction strength as normal, moderate, or weak.

Kappa = 0.44 (moderate)

Logistic Regression

Opposite of Linear regression, uses dichotomous measure of the outcome with the odds ratio. Predicts the value of one variable (dependent) based on one or more others (independent). Cause and effect

Example: Researchers are interested in whether the following factors influence the likelihood that a patient is discharged from the hospital to a long-term care facility versus their home. Factors included age, ADL status, marital status and gender

Statistical Measurement of Relationships consists of?

ICC >.75 is excellent and <.75 is moderate to poor

Cronbach's (.7-.9 is strong)

Parametric Tests of Differences (Independent)

Independent T-Test, One-Way Anova

Independent T-Test

What is the difference between two independent groups, t-statistic provided

Example: Therapist wants to determine whether a treatment was effective in reducing lower extremity edema in a group of patients with peripheral vascular disease. Volumetric measurements using water displacement is selected as the outcome measure (measured in milliliters). The data was compared to a control group receiving no treatment

One-Way ANOVA

Difference between two or more groups, f-statistic provided, post hoc needed

Example: A comparison of the effects of exercise in water, on land or combined (land + water) on the rehabilitation outcome of groups of patients with ACL reconstruction revealed that less joint effusion was noted after 8 weeks in the water group.

Non-parametric tests of differences (Independent)

Mann-Whitney U and Kruskal-Wallis

Mann Whitney U

Opposite of independent t-test, U score provided, difference between 2 independent groups

Example: Is there a difference in quadriceps strength (MMT score) between patients who received NMES-assisted therapy vs. standard care after ACL surgery?

Krusal-Wallis

Opposite of One-Way ANOVA, h-statistic provided, difference between two or more groups, post hoc required

Example: A study aims to investigate patient satisfaction (measured as unsatisfied, mildly unsatisfied, mildy satisfied, and satisfied) between three different groups (individual rehabilitation, group rehabilitation, and home exercise program).

Parametric Tests of Differences (Dependent)

Paired T-Test and Repeated ANOVA

Paired T-test

Difference within a group over time, t-statistic provided

Example: An investigator is examining the effect of a lower extremity plyometric program on baseline and post intervention drop jump performance (height in inches) in a group of subjects. Specifically, they want to know if the plyometric program significantly improves jump performance in this group of subjects.

Repeated Measures ANOVAs

Difference among repeated measures within 2 or more groups, f-statistic provided, requires post-hoc

Example: Researchers are interested in finding out the effect of multi-tasking on gait speed (in meters/sec) in a group of individuals with history of falls. They want to measure gait speed during 3 conditions: Walking at their preferred speed, walking while talking, and walking while holding a bag of groceries. Their hypothesis is that gait speed will be affected by multi-tasking

Non-Parametric Tests of Differences (Dependent)

Friedman's ANOVA and Wilcoxon Signed Rank

Wilcoxon Signed Rank

Opposite of paired T test, z-score provided

Example: Which of the following test would you use to assess whether there is a significant difference pre and post dry needling on a pain scale (severe pain, moderate pain, mild pain, and no pain) in individuals with low back pain?

Friedman's ANOVA

Opposite of Repeated ANOVA, f-statistic provided, requires post-hoc

Example: Do shoulder abductor MMT scores improve across three stages of outpatient physical therapy in patients with rotator cuff repair?

Chi Square (Non-Parametric)

Used in epidemiology studies, uses 2x2 table

Example: Is there an association between assistive device use and fall history in older adults?"