64 terms

From O'Sullivan& Siegelman 2009 NPTE guide

Systematic review, RCT, cohort study,homogeneity, case control study, case report

Types of Evidence and Research

Systematic Review (including meta-analysis)

a review in which the primary studies are summarized, critically appraised, and statistically combined; quantitative w/specific inclusion/exclusion criteria.

RCT (Randomized Controlled Trial)

experimental study, random sampling, random assignment to experimental or control group to receive different tx or placebo.

Cohort Study

prospective (forward in time) study; cohort w/similar condition is followed for defined period of time; comparison made to matched group w/out the condition.

Homogeneity

Systematic review free of variations in the directions and degree of results b/w individual studies.

Case Control Study

Retrospective (backward in time) study; group with similar condition compared with group that does not have condition to determine factors that might have played a role in the condition.

Case Report

Type of descriptive research in which only one individual is studied in depth, often retrospectively.

Levels of Evidence (best study design)

Level 1 (A), Level 2 (B) Level 3 (B) Level 4 (C)-Level 5(D)

Case series, poor quality cohort and case control studies. Descriptive

Level 4 (C)

Systematic review of cohort studies, prospective.

Level 2a (B)

Individual cohort study or low quality RCT (small N)

Level 2b (B)

Systematic review case controlled studies

Level 3a (B)

Individual case-control study, retrospective.

Level 3b (B)

Systematic review of multiple RCT's (large N)substantial agreement of size and direction of tx.

Level 1a (A)

Individual RCT w/narrow confidence level; tx effects precisely defined

Level 1b (A)

All or none case series. In absence of RCT, overwhelming evidence of substantial tx effect following intro of a new tx. (vaccine)

Level 1c (A)

Mean

average of all scores (X). Add all scores together and divide by the number of subjects (N). Used for interval and ratio data. Most common measure for central tendency.

median

midpoint, 50% of scores are above the mean and 50% are below, appropriate for ordinal data.

Mode

most frequently occurring score; used for nominal data.

Mean, Median, Mode

Measures of central tendency; determination of avg or typical scores.

Range, Standard Deviation, Normal distribution, Percentiles and quartiles

Measures of variability; determination of the spread of a groups scores

Range

difference b/w the highest score and lowest

SD (Standard Deviation)

determination of variability of scores (difference) from the mean. Subtract each score from the mean, square each difference, add up all the squares, and divide by the number of scores. For interval, ratio data.

Normal distribution

Symmetrical bell-shaped curve indicating the distribution of scores; the mean, median, and mode are similar. 1/2 the scores are above the mean and 1/2 below.

Normal distribution

Most scores are near the mean, within 1 SD; approx 68% of scores fall within +1 or -1 SD of mean.

Frequency of scores decreases further from the mean (Normal distribution)

95% of scores fall +2 or -2 SD of mean.

99% of scores fall +3 or -3 SD of the mean.

99% of scores fall +3 or -3 SD of the mean.

Percentiles

data divided in 100 equal parts

Quartiles

data divided into 4 equal parts and position of score is placed accordingly.

Inferential statistics

allow determination of how likely results of study of a sample can be applied to whole population.

Standard error of measurement

estimate of expected errors in individual's score; a measure of response stability or reliability.

Tests of significance

estimation of true differences, not due to chance; a rejection of the null hypothesis. ie. Probability levels or alpha levels.

Alpha levels (preselected level of statistical significance)

0.05 or 0.01; indicates expected difference is due to chance ie. 0.05, only 5 X out of 100 or 5% chance. P value. Allows rejection of null hypothesis; there are true differences on the measured DV.

Degrees of Freedom

based on # of subjects and # of groups; allows determination of level of significance based on consulting appropriate tables for each statistical test.

Standard Error

expected chance variation among the means, result of sampling error.

Type I Error

null hypothesis rejected by the researcher when it is true

Type II Error

null hypothesis is NOT rejected by the researcher when it is false.

Parametric statistics

based on population parameters; includes tests of significance based on interval and ratio data. Assume normal distribution in population, random sampling, variance in groups is equal.

T-test

parametric test compares 2 independent groups created by random assignment and identifies difference at a selected probability level ie 0.05.

T-test for independent sample, Paired samples; one tailed and two tailed T-tests.

Types of T tests

T-test for independent samples

compares the difference b/w 2 independent groups;

T-test for paired samples

compares difference b/w 2 matched samples; 2 subtests are one-tailed and two-tailed t-test.

One tailed t test

based on a directional hypothesis; evals difference in data only on 1 end of distribution, either negative or positive.

Two-tailed t-test

based on a nondirectional hypothesis; evals differences in data on both positive and negative ends of a distribution; tests of significance are almost always two-tailed.

Analysis of Variance (ANOVA)

parametric test too compare 3+ independent tx groups or conditions at a selected probability level.

Simple (one way) ANOVA

compares multiple groups on a single independent variable ie. 3 sets of posttest scores

Factorial ANOVA (multifactorial)

compares multiple groups on two or more independent variables.

Analysis of Covariance (ANCOVA)

parametric test used to compare 2 or more treatment groups or conditions while also controlling for the effects of intervening variables ie. subjects in one group taller than subjects in the second group-ht is covariate.

Nonparametric statistics

testing not based on population parameters; includes tests of significance based on ordinal or nominal data. Less powerful than parametric tests, more difficult to reject null hypothesis.

Chi square test

non-parametric test of significance used to compare data in the form of frequency counts occurring in 2 or more mutually exclusive categories ie. subjects asked to rate tx preferences.

Correlational statistics

to determine relative strength of a relationship b/w 2 variables.

Pearson product-moment coefficient (r)

used to correlate continuous data with underlying normal distribution on interval or ration scales. ie relationship b/w proximal and distal development in infants examined.

Spearman's rank correlation coefficient (rss)(rho)

nonparametric test to correlate ordinal data ie. verbal and reading scores

Types of Correlational statistical tests

Pearson-product-moment coefficient (r)

Spearman's rank correlation coefficient (rho), (rss)

Point biserial correlation

Rank biserial correlation

Intraclass correlation coefficient (ICC)

Strength of relationships

Common variance

Spearman's rank correlation coefficient (rho), (rss)

Point biserial correlation

Rank biserial correlation

Intraclass correlation coefficient (ICC)

Strength of relationships

Common variance

Point biserial correlation

one variable is dichotomous (nominal) and other is ratio or interval eg. the relationship b/w elbow flexor spasticity and side of stroke.

Rank biserial correlation

one variable is dichotomous (nominal) and other is ordinal ie. relationship b/w gender and functional ability

Intraclass correlation coefficient (ICC)

a reliability coefficient based on an analysis of variance

Strength of relationships

Positive correlations range 0-+1.00; indicates as variable X increases so does Y

High Correlations

>0.76 to +1.00

Moderate correlations

0.51 to 0.75

Fair correlations

0.26 to 0.50

Low correlations

0.00 to 0.25, 0 means no correlation

Negative correlations

range from -1.0 to 0, indicates as variable X increases, variable Y decreases; an inverse relationship.

Common variance

representation of the degree that variation in one variable is attributable to another variable. Determined by squaring the correlation coefficient ie. coefficient of .70 means common variance is 49%.

Linear regression

used to establish relationship b/w 2 variables as a basis for prediction. X is the independent variable or predictor, Y is the dependent or criterion variable. ie. Can SBP be predicted from age?