Term | Definition |
Alpha halves | The risk of Type I error when a test of equality is performed. If equality is tested, researchers assume either group could be superior,thus alpha divided in half. Test of equality referred to as a two-tailed test (there are two sides). |
Area under the curve (AUC) | The ratio of the false-positive rate to the true positive rate produces a proportion called the area under the curve. An AUC of 0.50 (50%) means the dx test is perfectly accurate. |
Coefficient of determination | r-squared (Pearson’s product-moment correlation coefficient). Tells the proportion of the variability in the response variable that is explained by the variability in the predictor variable. The higher the percentage, the stronger the association. |
Convenience sampling | Selecting the subjects of a study based on ease of access (convenience). |
Degrees of freedom (df) | (#observations - 1) |
Effect size | In stats power calculation; the expected amount of effect that is considered a minimal clinically meaningful effect. Often reported as a Cohen’s d statistic. |
Goodness-of-fit tests | A type of chi-square test that can test if the sample distribution fits with a theoretical distribution. |
Multistage sampleing | Procedure where sampling occurs in one more steps (stages). Often used when dealing with very large sampling frames and you want to ensure every subgroup is included. |
Nonprobability sample | Procedure in which subjects are purposefully selected in order to ensure they represent/experience the phenomenon on interest. Often used with rare conditions or small populations. |
One-sample t-test | Compares mean score of sample to known mean value from a trustworthy source. Central question: if there is difference between experimental group vs. known normative value. |
P-value | Probability that the test statistic represents normal variation. P-value reported w/each test statistic. P-value compared to lvl of alpha selected for study. If test statistic is less than alpha then result is deemed statistically significant. |
Pearson product-moment correlation coefficient | (r). Test for relationship between two variables that are interval or ratio data types. An r of at >0.755 is considered a strong positive association (-0.755 would be a strong negative association). AKA Pearson’s correlation coefficient. |
Power calculations | Determines the probability that the effect of interest will be observed when it occurs. 0.8 = sufficient level of power. Used in sample size calculation to determine necessary #subjects needed in order to attain desired level of power. |
Purposive sampling | Used when there are few cases of a specific problem, such as w/rare dx or when interested in experiences of specific subjects. Qualitative research. |
Regression analysis | Focuses on relationships among several variables that best explain specific response. Determines best fit line w/in set of coordinates. Goal: ID combo of predictor variables that best explain the variability in the response variable. |
Reliable | Consistency or reproducibility of manner in which the data were collected. |
Repeated measures test | Pre/post, w/in subject, self-controlled, and correlated group designs tests. Compares variables from different groups that don't influence one another,(independent), or when same subject in both groups, (dependent). |
Sampling fraction | The size of the sample as a percentage of the population from which it is drawn. |
Simple random sample | Form of probability sampling (each individual has known probability of being selected) hat allows every member of the population to have an equal chance of being selected for the study. |
Snowball sampling | Studying populations difficult to identify, such as (gang members, IV drug users, or illegal aliens). AKA “network sampling”. 1st subjects asked to ID others. Researchers then contact & determine if they fit. If good fit, add to sample, and repeated. |
Stratified random sample | Procedure where sampling frame is divided in to strata (groupings) and a random sample is drawn from each stratum. |
Test for differences | Comparison of a statistic, (mean or proportion) between two or more groups. |
Test statistic | Value produced by statistical test. |
Tests for relationships | Determines if variables change in relationship to other variables. (+) relationship = increase in one variable + increase in another. (-) relationship = increase in one variable is associated + decrease in another variable. More association = significant |
Truncated trials | A trial that ends early. Usually this happens for one of two reasons: (1) results indicate an unexpected adverse outcomes and the study is stopped for safety purposes; (2) results indicate a strongly positive outcome, making it necessary to continue. |
Valid | Accuracy of measurements. The degree to which and instrument measures what it is supposed to measure. |