Design and Independence
Use this section when you need to check whether pairing, repeated measures, batch structure, or analysis-unit choices could make observations non-independent.
Use this section when your question is whether observations are truly independent, whether the design should be treated as paired or repeated, or whether batch, plate, or analysis-unit structure could distort the inference.
This section is for checking design-level dependence risks before you over-interpret a result. It is not the right place to choose the primary statistical method or to evaluate general robustness to alternative specifications; if your first question is which test or model to run, start with Methods, and if your first question is about assumption sensitivity or outlier-driven fragility, start with Assumptions and Robustness.
How independence is handled
Independence is not handled by a single checker or a single formal test. Instead, related checks -- pairing, pseudoreplication, batch/plate confounding, temporal dependence, and analysis-unit concerns -- appear on separate surfaces.
Confirmations and disclosures are distributed across cards and snapshot state. Each check has its own confidence level and resolution state. When reading the child pages below, treat each as a separate surface rather than a component of one unified independence module.
Is this the right section?
Use the table below to choose the leaf page that best matches the design or independence question you need to answer.
| If your question is... | Start here |
|---|---|
| Should this comparison be treated as paired or unpaired? | Paired vs Unpaired Guard |
| Are multiple rows really independent, or could this be pseudoreplication? | Pseudoreplication Detection |
| Could batch or plate assignment be confounding the signal? | Batch and Plate Confounding Detection |
| Should repeated measurements be modeled as repeated measures rather than independent rows? | Repeated Measures Model Suggestion |
| Do I need a broader check on whether the unit of analysis is statistically independent? | Statistical Independence Check |
If your main task is choosing the analysis family itself rather than checking dependence structure, this is usually not your first stop. Use Methods before returning here.
Support boundary
This category covers design-level independence risks, not a standalone guarantee that the chosen method, assumptions, or conclusions are fully valid.
Child pages
- Paired vs Unpaired Guard
- Pseudoreplication Detection
- Batch and Plate Confounding Detection
- Repeated Measures Model Suggestion
- Statistical Independence Check
Start here
- Start with Paired vs Unpaired Guard if your first decision is whether observations are matched across conditions or timepoints.
- Continue to Pseudoreplication Detection if multiple rows may come from the same biological, technical, or experimental unit.
- Go to Batch and Plate Confounding Detection if assignment to runs, plates, or batches may be entangled with condition.
- Use Repeated Measures Model Suggestion when the same unit is measured more than once and should not be treated as independent rows.
- Use Statistical Independence Check when you need a broader check on whether the chosen analysis unit is independent enough for the planned inference.
Related
- Related links are registered in frontmatter and rendered below.