The accepted ontology
Browse accepted Systema Constructum entries: human-made tools, institutions, methods, practices, works, and their definitions.
100 entries in this page
Model capacity is a human-made measure of a machine-learning model's representational capability — the range and complexity of functions it can approximate given its architecture and parameters. Parameters: (1) parameter count (total trainable weights and bia…
A component-plus-residual plot (CPR plot) is a regression diagnostic visualization constructed by plotting each predictor variable against the quantity ŷ + ε̂, where ŷ is the model's fitted value and ε̂ is the residual. The plot overlays a lowess or LOESS smo…
Regression analysis is a human-made statistical methodology for estimating the functional relationship between a dependent variable and one or more independent variables. It operates through specified model forms (linear, generalized, nonparametric), estimati…
A leverage residual is a human-made diagnostic construct in linear regression analysis that combines residual magnitude with observation leverage to flag influential data points. It is computed as r_i * sqrt((1 - h_ii) / h_ii) where r_i is the raw residual fo…
Dfbetas is a regression diagnostic measure that computes the standardized change in each regression coefficient β_i resulting from the deletion of a single observation j from the fitting data. For a model with p predictors and n observations, dfbetas_ij = (β_…
A normal probability plot is a diagnostic visualization that maps the ordered residuals of a statistical model to the theoretical quantiles of a standard normal distribution, producing a Q-Q plot where systematic deviations from the reference line indicate de…
A diagnostic measure is a numerical statistic computed from a fitted statistical model and its data to evaluate model adequacy, flag problematic observations, or check assumption violations. Each measure isolates a specific aspect of model behavior — influenc…
An influential observation is a data point in a regression dataset whose removal would meaningfully alter the fitted model's coefficients or predictions. It is identified through diagnostic measures — Cook's distance, DFBETAS, dfits, and leverage — that quant…
Normalized residuals are regression diagnostics obtained by dividing each raw residual r_i by the square root of MSE times (1 - h_ii), where h_ii is the i-th diagonal element of the hat matrix and MSE is the mean squared error of the fitted model. Unlike raw…
Cook's distance is a regression diagnostic measure that quantifies the influence of each observation on the fitted regression coefficients. It is computed as D_i = (r_i^2 / (p * MSE)) * (h_ii / (1 - h_ii)^2) where r_i is the i-th standardized residual, h_ii i…
Regression diagnostic is a human-made analytical procedure that examines the residuals, influence, and specification errors of a fitted regression model to assess its reliability. Parameters: (1) residual pattern, (2) influence measures, (3) specification tes…
A partial leverage plot (also known as a component leverage plot) is a regression diagnostic visualization that decomposes the overall leverage of each observation into contributions from individual predictors. Each data point is assigned a partial leverage v…
A dfits measure is a regression diagnostic that quantifies the standardized difference between the fitted values from the full model and the fitted values from the model fit after deleting a single observation. For observation i, dfits_i = (y_hat_i - y_hat_i^…
A dfbeta is a regression diagnostic measure that quantifies how much each individual regression coefficient changes when a single observation is deleted from the model. For observation i and coefficient j, it is computed as dfbeta_j(i) = (beta_j - beta_j(i))…
A standardized residual is a statistical transformation that divides each residual (the difference between an observed value and its fitted value) by an estimate of its standard deviation. This standardization produces values that are approximately standard n…
A tolerance interval is a statistical construct for population-level uncertainty quantification. It specifies a range that, with a specified confidence level (1-alpha), contains at least a specified proportion (p) of the underlying population distribution. Pa…
Bootstrapping is a statistical resampling method that estimates the sampling distribution of a statistic (e.g. mean, median, regression coefficient) by repeatedly drawing samples with replacement from the observed dataset. Parameters: (1) B — the number of bo…
A studentized residual is a human-made statistical measure used in regression analysis to identify outliers. It adjusts raw residuals by dividing by an estimate of their standard deviation, accounting for the fact that residuals have non-constant variance. Th…
Busywork is a human-made organizational practice consisting of activities that signal effort or compliance while producing negligible substantive value. Parameters: (1) the activity is observable and reportable — it leaves a trace that can be presented as evi…
A partial regression plot (also called an added-variable plot) is a scatter plot that visualizes the marginal relationship between a single predictor and the response in a multiple regression model, after removing the linear effects of all other predictors. I…
A partial residual plot (also called a component-plus-residual plot) is a scatter plot used in multiple regression analysis to assess whether a predictor variable has a linear relationship with the response. It displays the partial residuals (observed respons…
A propensity score is the conditional probability of a unit receiving a treatment or exposure, given a set of observed covariates. It is computed by fitting a statistical model (typically logistic regression or a generalized linear model) to observed covariat…
A variance inflation factor is a scalar measure computed for each predictor in a multiple regression model. For predictor X_j, let R²_j be the coefficient of determination when X_j is regressed on all other predictors. The VIF is VIF_j = 1 / (1 - R²_j). The p…
An added-variable plot (also called a partial regression plot) is a regression diagnostic scatterplot that isolates the relationship between one predictor variable and the response by plotting the residuals of the response regressed on all other predictors ag…
A causal graph is a directed graph model where nodes represent variables and directed edges encode direct causal influences between them. Its parameters are: (1) a set of nodes V mapping to variables, (2) a set of directed edges E where (X,Y) ∈ E means X has…
A dataset is a structured collection of data records assembled by humans for use in analysis, modeling, or experimentation. It consists of observations (rows) described by variables (columns), curated and formatted so that computational tools can ingest and o…
A training set is a curated collection of labeled data instances used to adjust model parameters during supervised machine learning. It is defined by three parameters: (1) a matrix of feature vectors representing the training samples, (2) a corresponding vect…
A scale-location plot is a diagnostic scatterplot used in regression analysis to assess whether the variance of residuals is constant across fitted values. It maps each fitted value on the x-axis to the square root of the absolute standardized residuals on th…
A hat matrix is a human-made mathematical construct in linear regression analysis that maps observed response values to fitted values. It is an n×n projection matrix H = X(X'X)^{-1}X' where X is the design matrix, and its diagonal elements h_ii (leverage valu…
Cook's distance is a human-made statistical measure used in regression analysis to quantify the influence of each individual observation on the fitted regression coefficients. It carves a scalar value computed from the residual, leverage, and coefficient esti…
A sensitivity plot is a diagnostic visualization used in regression analysis to assess how residuals vary across the range of fitted or observed values. It typically displays residuals against one predictor or the fitted values, often with a smoothed trend li…
A shap plot is a human-made visualization technique in machine learning interpretability that displays SHAP (SHapley Additive exPlanations) values to quantify how each feature contributes to a model's prediction for individual instances. The mechanism of pers…
A leverage plot is a diagnostic visualization used in regression analysis to identify data points with high leverage — observations whose predictor values are unusual or extreme relative to the rest of the data. It plots each observation's leverage value (hat…
A diagnostic plot is a class of visualization used in statistical modeling to assess model adequacy by examining residuals, leverage, and influential observations. It transforms model output into structured visual form — typically scatterplots, Q-Q plots, or…
An empirical cumulative distribution function (ECDF) plot is a data visualization that displays the proportion (or count) of observations falling below each observed value, plotted as a step function. It shows the cumulative distribution of a single variable…
A network diagram is a graph-drawing that represents entities as nodes and their relationships as edges (directed or undirected), enabling visualization of topology, centrality measures, clusters, and connectivity patterns in complex systems. Developed from g…
A timeline is a linear representation of events ordered along a chronological axis. It organizes events, milestones, or states into a sequence indexed by time, enabling visualization of duration, simultaneity, and causal ordering. The parameters are: (1) a te…
A process-map is a visual diagram that represents the sequence of steps, decision points, inputs, outputs, and flow of activities within a process. It uses standardized symbols (rectangles for processes, diamonds for decisions, arrows for flow) to create a sh…
A structured, proactive risk-analysis method applied during the design phase of a product or system to identify potential failure modes, assess their effects on functionality and safety, and prioritize corrective actions before manufacture. It extends the gen…
A structured methodology for planning and conducting experiments to systematically study how input variables (factors) affect output responses. It operates through defined experimental designs — full factorial, fractional factorial, Latin square, and response…
A capability index is a numerical ratio that measures how well a process's natural variation fits within specified tolerance limits, calculated as the ratio of allowed spread to actual spread (e.g., Cp = (USL - LSL) / 6σ, Cpk adjusts for centering). It quanti…
An x-bar chart is a control chart that plots the arithmetic means of successive subgroups (samples) collected from a process, with center line at the overall mean and control limits at ±3 standard errors of the mean. It detects shifts in the process average b…
Process capability is a statistical measure of how well a manufacturing or business process can produce output within prescribed specification limits. It quantifies the relationship between the natural variation of a process and the tolerance envelope set by…
A run chart is a time-series line plot: a sequence of observed values plotted on the vertical axis against time or ordered observation index on the horizontal axis, with no control limits, used to visualise trends, shifts, and cycles in process performance. I…
Acceptance sampling is a statistical decision procedure for lot disposition based on inspecting a random sample of n items from a batch of size N and comparing the count of nonconforming units to a pre-defined acceptance number c. Its parameters are: the samp…
Statistical process control (SPC) is a quality management methodology that applies statistical methods to monitor and control a process in real time. It distinguishes common-cause variation (inherent to the system) from special-cause variation (assignable to…
Condition monitoring is a maintenance practice where the state of equipment is measured continuously or at regular intervals — using sensors, visual inspection, or performance data — to detect early signs of degradation, wear, or abnormal operation before fai…
A statistical graphic that combines a bar chart and a line graph: bars show individual values in descending frequency order, while the line shows the cumulative percentage across categories. It carves the 80/20 heuristic visually — the vital few bars that acc…
A disciplined maintenance strategy that determines which maintenance actions are required for each asset, prioritized by the criticality of failure modes, and schedules them by condition-monitoring data rather than fixed intervals. It carves maintenance as a…
Condition-based monitoring is a maintenance strategy that continuously or periodically measures one or more condition parameters of equipment to detect degradation early enough to schedule maintenance at the optimal time. It operates through a defined chain:…
A degradation model is a mathematical or conceptual framework that describes how the performance, integrity, or capability of a system, component, or material deteriorates over time, usage cycles, or environmental stress. It specifies the functional form of d…
system-availability quantifies the proportion of time a system remains operational within a defined period, expressed as a ratio (e.g., 99.9%) or a fraction of uptime versus total time. It is computed from uptime records maintained by monitoring infrastructur…
A quantitative reliability metric expressing the frequency of failures per unit of operation time, computed as the number of observed failures divided by the total accumulated operating time of the system or component. Parameters: (1) the population under obs…
A maintenance strategy in which upkeep tasks are performed on a fixed schedule (time-based or usage-based) rather than in response to failure or in response to observed degradation. The defining parameters are: (1) a predetermined interval — measured in calen…
Predictive-maintenance is a human-made practice for scheduling equipment upkeep based on actual condition data rather than fixed intervals. Its parameters are: (1) continuous or periodic monitoring of equipment state via sensors (vibration, temperature, acous…
A hazard-and-operability study (HAZOP) is a structured, systematic risk assessment methodology that identifies potential hazards and operability deviations in industrial processes. Its parameters are: (1) guide words (NO FLOW, MORE PRESSURE, LESS TEMPERATURE,…
A service mesh is a dedicated infrastructure layer for managing service-to-service communication in distributed systems. Its parameters define two deployment modes: the data plane (sidecar or ambient proxies intercepting all network traffic between services)…
Human-error is a systematic classification of error modes attributed to human operators within engineered or organizational systems. Its parameters distinguish between active failures (slips, lapses, mistakes, violations) and latent conditions (organizational…
A digital twin is a dynamic computational replica of a physical entity — a device, process, or system — that mirrors its real-time state through continuous data ingestion from sensors or telemetry. The twin runs simulations or analytical models to support pre…
A cyber-physical-system is a human-engineered architecture that tightly integrates computational algorithms, communication networks, and physical processes through continuous bidirectional feedback loops. Its parameters are: (1) cyber elements that perform se…
Fault injection is a testing methodology in which failures are intentionally introduced into a system's normal operation to observe degradation behavior, contain failure modes, and validate recovery mechanisms. The technique persists through standardized fram…
A symbolic system is a human-made representational construct consisting of a finite inventory of symbols paired with rules for combining those symbols into well-formed expressions. The symbols themselves are arbitrary marks or sounds; their meaning arises ent…
Disaster-recovery is the human-made discipline of planning, preparing, and executing the restoration of IT systems, data, and business operations after major disruptions. It is structured around procedures, protocols, strategies, and technologies designed to…
Recovery-time-objective (RTO) is a disaster recovery parameter specifying the maximum acceptable duration of service disruption after a failure event. It defines the target time by which systems, processes, or data must be restored following an incident. RTO…
Mean-time-to-failure (MTTF) is a reliability metric that quantifies the average elapsed time from deployment or activation to the first failure of a non-repairable component or system. It is calculated as total operational time divided by the number of failed…
A recovery-point-objective (RPO) is a time-bound parameter that specifies the maximum tolerable data loss for a system or process, expressed as the elapsed time between the last successful data backup and the failure event. It is established through business…
Meantime between failures (MTBF) is a reliability metric that quantifies the average elapsed time between successive failures of a repairable system. It is calculated as total operational time divided by the number of failures over a given period. MTBF serves…
A reliability block diagram (RBD) is a graphical model that represents a system's reliability structure using numbered blocks connected by lines. Each block encodes a component's failure rate or reliability function; the interconnections encode how component…
A chaos-monkey is a tool that randomly and autonomously injects faults (killing services, introducing latency, corrupting data, or removing resources) into a running production environment. It is characterized by: (1) randomness — faults are injected unpredic…
A chaos-experiment is a controlled intervention that injects faults, outages, or degradation into a running system to verify that resilience mechanisms operate correctly under stress. It is characterized by: (1) a specific hypothesis about system behavior und…
Monitoring is the systematic practice of observing, measuring, and tracking system behavior, operational metrics, and service indicators to maintain awareness of current state, detect deviations from expected behavior, and enable timely intervention. Paramete…
Blameless culture is an organizational practice in which incident responses and post-incident reviews focus on identifying systemic failures and improving processes rather than assigning individual culpability. Parameters: (1) the practice replaces punitive f…
An incident-command-post is an incident coordination point — a designated physical location or virtual channel where incident response teams convene during active incidents to coordinate actions, track status, and communicate updates. Parameters: (1) establis…
A war room is a dedicated physical or virtual space assembled to coordinate high-stakes operations — incident response, crisis management, or time-critical decision-making. It concentrates expertise, communication channels, and decision authority in one locus…
Error-injection is a testing technique in which faults, failures, or anomalous conditions are deliberately introduced into a system to verify that it detects, handles, and recovers from errors gracefully. Parameters: (1) the fault type injected (e.g., network…
An on-call is an operational staffing practice in which a designated person or rotation of people remain available outside normal working hours to respond to incidents, alerts, or service disruptions when they occur. Its parameters include the on-call rotatio…
An alert-threshold is a numeric or categorical value configured in a monitoring system that, when exceeded or crossed by a measured signal, triggers an alert condition. Its parameters include the signal identifier, the threshold value, the comparison operator…
A health-check is a monitoring probe that tests whether a system, service, or component is operating within acceptable parameters by sending a minimal request and evaluating the response. Its parameters include the probe type (TCP connect, HTTP request, scrip…
A post-incident-review (PIR) is a structured organizational practice that examines a completed incident through four phases: (1) timeline reconstruction using available logs and records, (2) root-cause analysis distinguishing immediate triggers from systemic…
Failure-mode-effects-analysis (FMEA) is a structured, proactive reliability-engineering method that: (1) decomposes a system, component, or process into its constituent functions; (2) identifies every possible failure mode for each function — how it could fai…
Toil is the class of operational tasks that are (1) manual and repetitive in nature, (2) tied to a specific service iteration rather than enduring, (3) scalable away with automation, (4) not intended to permanently improve the service's long-term health, and…
A degradation-mode is a human-made classification of failure behaviors where a system responds to a fault by reducing its capability proportionally rather than ceasing operation entirely. Its parameters are: (1) the fault that triggers degradation, (2) the de…
An incident-command system is a human-made organizational practice that structures the coordination of incident response by assigning a dedicated incident commander with authority over all response activities, supported by role-holders (communications lead, t…
An alerting system is a human-made notification infrastructure that maps monitored conditions to prioritized signals routed to responsible operators. Its parameters are: (1) alert rules that define thresholds or anomaly-detection criteria for triggering, (2)…
An alarm is a human-made signaling artifact that interrupts normal operation to draw attention to a specific condition — danger, anomaly, or required action — through one or more sensory channels (auditory, visual, tactile, or textual). Its parameters are: (1…
Maintainability is the degree to which a system, component, or artifact can be maintained — repaired, inspected, adjusted, or updated — within specified conditions, time constraints, and resource limits. It is quantified through metrics such as mean time to r…
A safety survey is a systematic evaluation of safety conditions, conducted through structured observation, checklist review, and interviews, to identify hazards, assess risks, and verify compliance with established safety standards and regulations. It persist…
MTBF (Mean Time Between Failures) is a reliability engineering metric that measures the average time between consecutive failures of a repairable system. It is calculated as the total operational time divided by the number of failures observed during that per…
mean-time-to-repair is a reliability metric that quantifies the average duration from the moment a system component fails to the moment it is fully restored to operational status. It is calculated by dividing total maintenance downtime by the number of failur…
Systematic-inspection is a structured practice in which human-made checklists, procedures, and scheduled routines are applied to examine systems, processes, or components in a methodical and repeatable manner. Its parameters include: (1) pre-defined scope and…
A common-cause failure (CCF) is a safety engineering construct in which multiple redundant components or safety channels fail simultaneously due to a shared root cause that circumvents the independence assumed by their redundancy. Parameters: (1) two or more…
Defense-in-depth is a security and reliability engineering construct in which multiple, independent layers of protection or failure mitigation are deployed so that the compromise or failure of any single layer does not lead to system breach or collapse. Its p…
A failure-analysis is a systematic methodology for identifying, investigating, and characterizing the root causes and failure mechanisms of components or systems after a failure event. Its parameters are: (1) a failure event is observed, (2) evidence is colle…
A fail-soft system is one designed to maintain partial or reduced operation when a component or subsystem fails, rather than ceasing operation entirely. Its parameters are: (1) a failure mode is identified, (2) fallback pathways or reduced-capacity modes are…
A latent-condition is an analytical construct used in safety engineering to categorize hidden flaws, systemic deficiencies, or organizational weaknesses embedded within a technical or management system — design oversights, procedural gaps, training deficienci…
A safety-instrument is a mechanism designed to detect an unsafe condition or deviation from safe operating parameters and execute an action that brings the system into a safe state. Its parameters are: (1) sensing — the instrument monitors one or more process…
A single-point-of-failure is a structural weakness in a system design where the failure of one component causes the entire system to cease functioning. Its parameters are: (1) uniqueness — exactly one component performs the critical function; (2) cascade — th…
Defensive design is a design approach that anticipates likely failure modes and structures systems to handle errors gracefully rather than catastrophically. Its parameters are: (1) failure anticipation — identifying probable failure modes before they manifest…
Blast-radius is a conceptual construct that defines and bounds the scope of damage produced by a single point of failure within a system. Its parameters are: (1) failure mode identification — specifying which component or layer has failed; (2) dependency mapp…
A structured procedure for transferring operational responsibility, critical information, and contextual awareness between outgoing and incoming personnel at defined shift boundaries. Parameters: (1) a designated outgoing operator who holds current state know…