1/21
Vocabulary flashcards covering the Neyman-Pearsonian framework for Null Hypothesis Statistical Testing (NHST), Type I and Type II errors, test statistics, directional and non-directional decision rules, and confidence intervals.

David Bakan
Psychologist who characterized the test of statistical significance in psychological research as an instance of a kind of essential mindlessness in the conduct of research.
Steps of NHST
A seven-step procedure: (1) State the population distribution of the random variable(s), (2) State the null and alternative hypotheses, (3) State the assumptions of the test, (4) Identify the test statistic and its sampling distribution, (5) Specify alpha and a decision rule, (6) Conduct the study, calculate descriptive statistics, and check assumptions, and (7) Run the test and make a decision.
Central Limit Theorem (in NHST)
The theorem establishing that if a parent population distribution is X∽N(Xˉˉ,NXˉˉ), the sampling distribution of the sample mean is XˉN∽N(XˉˉXˉ,NXˉˉ2), allowing researchers to deduce expected values despite lacking direct access to the parent population.
Research Hypothesis
A statement that describes the general aim or substantive prediction of the researcher.
Null Hypothesis (H0)
A testable claim specifying what is expected of the parameter(s) of interest when there is no interesting effect (e.g., H0:Xˉˉ=3).
Alternative Hypothesis (H1)
The complement of H0 that covers the outcome where an interesting effect exists and the research hypothesis is supported (e.g., H1:Xˉˉ=3).
Non-Directional Hypotheses
Hypotheses that test for any difference without specifying direction (e.g., H0:Xˉˉ=3 vs. H1:Xˉˉ=3), evaluated using a two-tailed test.
Directional Hypotheses
Hypotheses specifying the expected direction of the parameter difference (e.g., H0:XˉˉXˉˉ3 vs. H1:Xˉˉ<3 or H0:XˉˉXˉˉ3 vs. H1:Xˉˉ>3), evaluated using a one-tailed test.
Type I Error
The error committed when rejecting a null hypothesis (H0) that is actually true, with a long-run probability defined as Xˉˉ=P(rejecting H0XˉˉH0 True).
Type II Error
The error committed when retaining/failing to reject a null hypothesis (H0) that is actually false, with a probability denoted as Xˉˉ=P(type II error).
Neyman-Pearson Decision Matrix
The four possible binary decision outcomes: retaining a true H0 (correct), rejecting a false H0 (correct), rejecting a true H0 (Type I error with probability Xˉˉ), or retaining a false H0 (Type II error with probability Xˉˉ).

Test Statistic (statobs)
A statistic calculated from sample data that is used to make an inference about an unknown population parameter.
Null Distribution
The sampling distribution of the test statistic or sample mean under the condition that the null hypothesis (H0) is true.
Standardized Test Statistic (zobs)
The standardized test statistic obtained via linear transformation using the Central Limit Theorem: zobs=XˉˉNXˉˉXˉ−Xˉˉ∽N(0,1).
Observed p-value (pobs)
The probability of obtaining a test statistic value as extreme as or more extreme than the observed statistic, calculated under the null distribution.
Alpha (Xˉˉ) Cutoff
The pre-specified threshold probability (such as 0.05) chosen to define the region of values considered improbable under the null distribution, serving as the maximum acceptable Type I error rate.
Critical Value (statcrit)
The specific boundary value or threshold associated with the cutoff specified by alpha (Xˉˉ), such as zcrit=Xˉˉ1.65 for a one-tailed test at Xˉˉ=0.05, or zcrit=Xˉˉ1.96 for a two-tailed test at Xˉˉ=0.05.
NHST Decision Rules
Standard decision criteria: reject H0 if pobs<Xˉˉ or if statobs is more extreme than statcrit; otherwise retain/accept H0.
Point Estimate
A single sample statistic (e.g., Xˉ, s2, s) calculated to provide a single best guess for a population parameter.
Confidence Interval ((1−Xˉˉ)%)
An interval estimate constructed around a sample statistic such that over infinite random sampling of size N, (1−Xˉˉ)% of the calculated intervals will capture the true population parameter.

Confidence Limits (UCL and LCL)
The upper and lower numerical boundaries of a two-sided confidence interval, calculated as UCL=Xˉ+∣zcritical∣×SE and LCL=Xˉ−∣zcritical∣×SE.
Confidence Interval Decision Rule
The decision criterion stating that if the confidence interval built around the sample mean captures the value of Xˉˉ asserted by the null hypothesis, accept/retain H0; otherwise, reject H0.