Core scientific concepts, biases, and experimental design principles
25 cards · science
Sign up to start studying this deck
| Front | Back |
|---|---|
| Observation | Noticing phenomena that prompt a question Careful, systematic noticing of events or patterns starts inquiry. |
| Question | A focused query about an observed phenomenon Defines what you aim to investigate or explain. |
| Hypothesis | Testable explanatory statement about a phenomenon Must be falsifiable and lead to specific predictions. |
| Experiment | A controlled test of a hypothesis Manipulates variables to assess causal effects. |
| Data Analysis | Interpreting data to evaluate the hypothesis Uses statistical methods to separate signal from noise. |
| Conclusion | Judgement on whether results support the hypothesis May accept, reject, or refine the hypothesis based on evidence. |
| Replication | Repeating a study to verify results Confirms reliability and guards against false positives. |
| Randomization | Random assignment to reduce confounding Helps balance known and unknown factors across groups. |
| Control Group | Comparison group not receiving the intervention Isolates the effect of the treatment by providing a baseline. |
| Blinding | Concealing group assignment from participants or staff Reduces bias in behavior, reporting, and measurement. |
| Double-Blind | Both participants and researchers are blinded Minimizes placebo and observer effects simultaneously. |
| Confounding Variable | Factor related to exposure and outcome that distorts effect Can mimic or mask true causal relationships if uncontrolled. |
| Statistical Power | Probability of detecting a true effect Higher with larger samples, larger effects, and lower noise. |
| Internal Validity | Degree a study establishes a causal relationship Strengthened by control, randomization, blinding, and compliance. |
| Randomized Controlled Trial | Experiment with random assignment to intervention vs control Often considered the gold standard for causal inference. |
| Cohort Study | Observational study following groups over time by exposure Estimates incidence and relative risk; can be prospective or retrospective. |
| Case-Control Study | Observational study comparing prior exposures in cases vs controls Efficient for rare outcomes; typically reports odds ratios. |
| p-value | Probability of data as extreme if null is true Not the probability the null is true; smaller suggests less compatibility. |
| Confidence Interval | Range of values compatible with the data at a given level A 95% interval expresses uncertainty around an estimate. |
| Type I Error | Incorrectly rejecting a true null hypothesis A false positive; the long-run rate is alpha (e.g., 0.05). |
| Type II Error | Failing to reject a false null hypothesis A false negative; probability is beta; power is 1 − beta. |
| Correlation vs Causation | Association does not imply cause-and-effect Confounding, bias, or chance can produce spurious correlations. |
| Confirmation Bias | Favoring information that supports preexisting beliefs Example: reading only news that aligns with your views. |
| Survivorship Bias | Focusing on successes while ignoring failures Example: studying only companies that lasted skews conclusions. |
| Selection Bias | Systematic differences in how participants are chosen Skewed samples misrepresent the target population or effect. |