TASC Student Directed Inquiry cheat sheet

A researchable question names a specific population, variable, and measurable outcome; vague topics ("pollution") are not questions until narrowed.

Closed questions yield categorical or numeric answers (yes/no, a count); open questions yield explanatory or descriptive answers.

A testable hypothesis states a predicted relationship between an independent and dependent variable and must be falsifiable.

Operational definitions convert abstract constructs into measurable procedures (e.g., "engagement" = minutes on task per class period).

Scope check: a question is feasible if data can be gathered within the available time, tools, ethics, and access constraints.

Causal questions require manipulation or strong observational design; correlational questions ask only whether variables co-vary.

A question is non-researchable when the answer is settled fact, purely opinion, or unobservable with any available method.

Refining loop: broad topic to narrowed focus to draft question to variable identification to feasibility test to final question.

Primary sources report original data or first-hand accounts; secondary sources summarize or interpret others' work.

Peer-reviewed journal articles carry more evidentiary weight than news summaries, blogs, or commercial pages.

Currency, relevance, authority, accuracy, and purpose (CRAAP) form a checklist for judging any source.

Corroboration rule: a claim is stronger when at least two independent sources agree; a single source supports only a tentative statement.

Bias indicators include loaded language, missing citations, undisclosed funding, and one-sided selection of evidence.

Citation must allow a reader to relocate the source: author, date, title, and publisher or DOI/URL with access date.

Plagiarism includes paraphrasing without attribution, not only verbatim copying; quotations need both marks and citation.

Sample bias check: ask who was included, who was excluded, and whether the sample matches the population the claim covers.

Controlled experiment: manipulate the independent variable, measure the dependent variable, hold all other variables constant.

Control group receives no treatment or a placebo; the experimental group receives the treatment, isolating the variable's effect.

Random assignment to groups reduces selection bias; random selection from a population supports generalization.

Confounding variable changes alongside the independent variable and offers a rival explanation for the result.

Blinding hides treatment status from participants; double-blinding also hides it from data collectors to limit expectancy effects.

Sample size affects precision: larger samples shrink random error, but do not fix a biased sampling method.

Validity types: internal (causal inference sound), external (generalizes), construct (measures the intended concept).

Reliability means repeated measurement gives consistent results; a measure can be reliable yet invalid.

Ethics: informed consent, voluntary participation, confidentiality, minimal harm, and review approval where required.

Quantitative data are numeric and analyzed statistically; qualitative data are descriptive and analyzed thematically.

Mean is sensitive to outliers; median resists them; mode identifies the most frequent value and suits categorical data.

Standard deviation measures spread around the mean; a large value signals high variability among observations.

Correlation coefficient r ranges from -1 to +1; r near 0 means no linear relationship, not proof of no relationship.

Correlation does not establish causation; a third variable or reverse direction may explain the association.

Percent change = (new - old) / old x 100; percentage points describe a difference between two percentages.

Graph choice: bar charts for categorical comparisons, histograms for distribution shape, line graphs for trends over time.

Outlier handling requires a stated rule applied consistently, and any exclusion must be reported with the reason.

Data quality checks: consistent units, calibration of instruments, duplicate entries, and missing-value documentation.

A claim is the position asserted; evidence is the data supporting it; reasoning explains why that data supports the claim.

A warranted conclusion stays within the limits of the design, sample, and measured variables.

Alternative explanation: before accepting a causal claim, rule out confounds, selection effects, and measurement error.

Statistical significance is not practical importance; effect size and context determine whether a result matters.

Anecdote and single cases are weak evidence because they lack comparison groups and are subject to selection.

Counterargument handling strengthens a case: state the opposing view, then rebut with evidence or concede its scope.

Overgeneralization occurs when findings from a narrow sample are extended to a broader population without justification.

Logical fallacies to screen out: false cause, hasty generalization, appeal to authority, and cherry-picked evidence.

Standard report order: question and context, methods, results, interpretation, limitations, conclusion, references.

Results report what was found (numbers, patterns); discussion explains what it means and how confident the claim is.

Every figure needs a title, labeled axes with units, a legend when multiple series appear, and a caption stating the takeaway.

Limitations section names specific constraints (sample size, design, measurement) and their likely direction of effect.

Audience adaptation governs vocabulary, amount of background, and depth of statistics presented.

Oral presentation timing rule: roughly one minute per slide; put full detail in the narrative, not dense slide text.

Data visualization integrity: avoid truncated axes, inconsistent scales, and chart types that exaggerate differences.

Citation style consistency matters more than which style is used; every in-text citation must appear in the reference list.