Statistical Data Analysis and Reporting Training Course

Participants can write and run basic R code to inspect and manipulate data objects.

• Vectors, matrices, lists and data frames used in R analysis
• Assigning objects and inspecting structures with str, class and glimpse
• Atomic data types, missing values and factor variables
• Functions, arguments and return values in readable analysis scripts
• Logical operators and conditional statements for analytical rules
• Indexing rows, columns and subsets using base R syntax
• Installing packages and loading libraries with dependency awareness

Participants can import common data files and profile their structure before analysis.

• CSV, Excel and delimited files imported with readr and readxl
• Character encodings, date formats and locale settings checked during import
• Column names standardised using janitor naming conventions
• Data dictionaries matched against imported fields and business definitions
• Missingness profiles calculated by field, record and reporting period
• Duplicate records identified using keys, composites and transaction identifiers
• Initial profiling tables produced with summary, skimr and count

Participants can clean, recode and transform analysis datasets using dplyr procedures.

• dplyr verbs filter, select, mutate, arrange and summarise
• Case_when rules used to recode categories and derived flags
• Date calculations created with lubridate for periods and durations
• Outlier treatment documented using business rules and distributional evidence
• Missing values handled through deletion, imputation and explicit categories
• Joins checked using primary keys, cardinality and unmatched records
• Grouped summaries created for segments, periods and operational units

Participants can reshape and combine datasets into analysis-ready structures.

• pivot_longer converts repeated measure columns into observation records
• pivot_wider creates reporting tables from long-format analytical data
• Relational joins selected through one-to-one and one-to-many relationships
• Crosswalk tables reconcile changing codes, labels and organisational structures
• Aggregations aligned with transaction, customer, account and monthly grains
• Unit-of-analysis errors detected before calculating summary measures
• Derived datasets saved with documented transformation steps and dates

Participants can investigate distributions, relationships and anomalies before formal modelling.

• Frequency tables reveal category imbalance, rare values and coding errors
• Histograms, density plots and boxplots assess numerical distributions
• Scatterplots examine relationships, clusters and influential observations
• Faceting compares distributions across business units, cohorts and periods
• Correlation matrices screen numerical variables before regression analysis
• Exploratory questions distinguish signals, anomalies and data-collection issues
• Analysis notes record observations, hypotheses and required follow-up checks

Participants can create clear charts that accurately communicate statistical findings.

• ggplot2 grammar maps variables to aesthetics, geoms and scales
• Bar charts distinguish counts, proportions, rates and grouped comparisons
• Line charts display trends with appropriate dates, intervals and baselines
• Confidence intervals shown using error bars and uncertainty ribbons
• Colour palettes selected for accessibility, contrast and categorical meaning
• Annotations identify targets, thresholds, events and material changes
• Chart titles state findings rather than merely naming variables

Participants can calculate and interpret descriptive measures and sampling characteristics.

• Mean, median and mode compared for skewed operational data
• Variance, standard deviation and interquartile range quantify spread
• Percentiles and quantiles identify thresholds, service levels and tails
• Sampling frames, selection bias and nonresponse affect inference quality
• Central Limit Theorem supports approximate sampling distributions
• Standard errors distinguish estimate precision from underlying variation
• Confidence intervals reported with estimates, units and decision context

Participants can select, conduct and explain common statistical hypothesis tests.

• Null and alternative hypotheses linked to measurable decision questions
• P-values interpreted alongside effect sizes and confidence intervals
• One-sample and two-sample t-tests compare numerical group outcomes
• Paired t-tests assess before-and-after changes for matched records
• Chi-square tests examine associations between categorical variables
• ANOVA compares means across three or more independent groups
• Nonparametric tests selected when normality or scale assumptions fail

Participants can check test assumptions and communicate results without overstating conclusions.

• Normality assessed through Q-Q plots, histograms and Shapiro-Wilk tests
• Equal variance examined using residual plots and Levene's test
• Independence assessed from study design, timing and data-generation processes
• Multiple testing risk controlled using Bonferroni and false-discovery adjustments
• Effect sizes reported using Cohen's d, eta-squared and odds ratios
• Statistical significance separated from operational and policy significance
• Limitations stated for sample quality, assumptions and causal claims

Participants can build, interpret and assess linear regression models for continuous outcomes.

• Ordinary least squares models estimate relationships with continuous outcomes
• Model formulas specify dependent variables, predictors and interaction terms
• Coefficients interpreted with units, reference categories and holding-constant logic
• Adjusted R-squared and residual standard error assess model fit
• Residual plots check linearity, variance patterns and normality
• Influential records identified through leverage, Cook's distance and residuals
• Prediction intervals distinguished from confidence intervals for mean estimates

Participants can model binary outcomes and assess classification performance.

• Logistic regression models probabilities for binary outcome variables
• Odds ratios translated into meaningful changes in likelihood
• Reference groups selected deliberately for categorical predictor interpretation
• Predicted probabilities calculated for defined customer or case profiles
• Confusion matrices summarise true positives, false positives and errors
• Sensitivity, specificity, precision and recall assess classification choices
• ROC curves and AUC compare binary classification model discrimination

Participants can analyse time-based patterns and create basic statistical forecasts.

• Time-series structure separates trend, seasonality, cycles and irregular variation
• Date indexes checked for gaps, duplicated periods and aggregation errors
• Moving averages smooth volatile operational and financial series
• Seasonal decomposition identifies recurring patterns within historical data
• Autocorrelation plots reveal dependence between adjacent time periods
• ARIMA model components selected using differencing and correlation evidence
• MAE, RMSE and MAPE compare forecast accuracy

Participants can use multivariate methods to identify patterns across several variables.

• Standardisation prepares variables measured on different numerical scales
• Principal component analysis reduces correlated variables into interpretable components
• K-means clustering groups records using distance-based similarity measures
• Cluster counts selected using within-group variation and business usability
• Segment profiles compare size, value, risk and behavioural characteristics
• Correlation and multicollinearity checked before multivariate model interpretation
• Multivariate findings validated against domain knowledge and operational feasibility

Participants can make analytical work reproducible, reviewable and easier to maintain.

• Reproducible scripts run from raw data to final outputs
• Project folders separate source data, code, outputs and documentation
• Functions remove repeated code from recurring analytical tasks
• Parameters support repeatable reports for periods, regions and portfolios
• Data validation checks test ranges, uniqueness, completeness and referential integrity
• Peer review checks logic, assumptions, outputs and written conclusions
• Session information records package versions and R environment details

Participants can produce structured statistical reports that combine code, evidence and interpretation.

• Quarto documents combine narrative, code, tables and figures
• YAML metadata controls report titles, formats and output settings
• Code chunks manage visibility, messages, warnings and evaluated results
• Inline R code inserts current statistics into written findings
• gt tables format estimates, intervals, units and explanatory notes
• Report sections separate methods, findings, limitations and recommendations
• Rendered HTML and PDF outputs prepared for stakeholder distribution

Participants can translate analytical results into clear findings for non-technical stakeholders.

• Executive summaries lead with decisions, findings, implications and actions
• Results narratives state magnitude, uncertainty, comparison and business meaning
• Technical detail placed in appendices without hiding material limitations
• Visual hierarchy directs attention to exceptions, drivers and thresholds
• Recommendations distinguish evidence, judgement, dependencies and implementation owners
• Stakeholder questions anticipated through assumptions, scenarios and sensitivity analysis
• Plain-language explanations avoid causal claims unsupported by observational data

Participants can apply governance and ethical controls to statistical analysis and reporting.

• Personal data minimisation limits fields collected, retained and reported
• Anonymisation risks assessed through direct identifiers and re-identification combinations
• Small-cell suppression reduces disclosure risk in published tables
• Data lineage records source systems, transformations, owners and approval points
• Bias checks examine representation, measurement errors and proxy variables
• Purpose limitation links analysis activities to approved business questions
• Retention and access rules protect analytical datasets and outputs

Participants can review analysis critically and defend statistical choices before publication.

• Review questions test data suitability, method choice and result interpretation
• Sensitivity analysis examines conclusions under alternative assumptions and exclusions
• Scenario comparisons show effects of changed thresholds and input values
• Reconciliation checks align reported totals with approved source figures
• Peer challenge distinguishes correlation, confounding and credible causal evidence
• Decision logs record approvals, unresolved issues and reporting caveats
• Presentation structure handles technical scrutiny and executive time constraints

Participants can build, review and present a complete R-based statistical analysis and decision report.

• Business case translated into analytical objectives, measures and decision questions
• Source dataset profiled, cleaned and documented using reproducible R scripts
• Exploratory findings supported by appropriate tables, charts and summaries
• Statistical tests or models selected with stated assumptions and limitations
• Quarto report produced with methods, results, recommendations and appendices
• Peer review applied to code, calculations, charts and written claims
• Final presentation defends findings, uncertainty and proposed management actions