616 templates · page 7 of 7
Scientific Plots diagram templates
Scatter, line, heatmap, box, violin, density, contour, regression.
Remote sensing — radiance to validated surface product
Trace a satellite observation through radiometric calibration, atmospheric correction and geolocation. Compare retrieved surface quantities with independent observations while preserving cloud and quality masks.
ScienceBayesian workflow — prior predictive to posterior checks
Organise a Bayesian analysis around a generative model, prior predictive checks, computation diagnostics and posterior predictive comparison. Sensitivity to prior and likelihood choices is part of the result.
ScienceCausal inference — estimand, graph and identification
Separate the causal question from an estimable statistical association. Specify temporal ordering and assumptions before selecting adjustment variables; conditioning on a collider can introduce bias.
ScienceConstrained optimisation — feasibility and optimality
Organise a smooth constrained optimisation problem around objective, constraints, feasible candidates and optimality checks. KKT conditions require suitable constraint qualifications and are not sufficient for arbitrary non-convex problems.
ScienceNumerical solution — consistency, stability, convergence
Plan a numerical verification study using a reference problem, mesh refinement and error norms. A visually smooth solution is not evidence of convergence; stability and implementation checks address different risks.
ScienceMarkov model — states, transitions and validation
Specify mutually exclusive states and a transition interval before estimating probabilities. Rows of the transition matrix must sum to one; homogeneity and memorylessness are assumptions to test rather than automatic properties of the data.
ScienceUncertainty propagation — inputs and dependence
Trace a derived quantity to measured inputs, sensitivity coefficients and covariance assumptions. Independent-input root-sum-squares is a special case; correlated uncertainties require covariance terms.
ScienceMissing data — mechanism and sensitivity
Describe missingness patterns before choosing an analysis. Missing-at-random assumptions cannot generally be proven from observed data; sensitivity analyses address departures from the primary assumptions.
ScienceNetwork analysis — topology versus measurement
Separate a true conceptual network from the graph observed through a sampling process. Missing nodes and edges alter centrality and communities; layout proximity is not a measured relationship.
ScienceSequential experiments — stopping and error control
Plan repeated looks at accumulating data before an experiment starts. The stopping rule, decision thresholds and multiplicity strategy belong to the design; repeatedly applying a fixed-sample threshold changes error rates.
ScienceFactorial experiment — main effects and interactions
Design a two-factor experiment with every factor combination represented and replication allocated at the experimental-unit level. An interaction means the effect of one factor depends on the level of another.
SciencePredictive modelling — split before learning
Preserve an untouched evaluation set and fit preprocessing only within training folds. Feature selection, imputation and tuning can leak outcome information if performed before splitting.
ScienceDynamical systems — equilibria and local stability
Analyse a smooth dynamical model through equilibrium solutions and a local linear approximation. Local eigenvalue behaviour does not by itself establish global stability or rule out nonlinear phenomena.
ScienceReaction–diffusion model — kinetics and transport
Separate local reaction kinetics from spatial diffusion and boundary fluxes. Spatial pattern predictions depend on parameters, initial conditions and numerical resolution; a pattern image alone does not establish mechanism.
ScienceMultiscale modelling — coupling and validation
Connect material-scale parameters, component behaviour and system predictions without hiding scale transitions. Calibrate and validate with separate evidence and propagate parameter uncertainty across the coupling.
ScienceAcoustic measurement — source, room and sensor
Distinguish the source spectrum from room propagation and microphone response. Background noise, weighting and averaging settings must be recorded for meaningful comparisons.
Science