Quantitative Data Analysis and Visualisation with Python Training Course

Participants can manipulate arrays and indexed data structures for quantitative calculations.

• NumPy arrays created from lists, ranges and imported numerical values
• Array indexing, slicing and Boolean masks used for record selection
• Vectorised calculations replacing slow row-by-row Python processing
• Broadcasting rules applied to standardise and transform numerical measures
• Pandas Series constructed with values, labels and missing observations
• DataFrame columns selected, assigned and calculated using explicit expressions

Participants can import common data files and establish an initial dataset profile.

• CSV, Excel and JSON files imported with appropriate parsing options
• Column names standardised using documented naming and formatting rules
• Data types inspected before calculation, grouping or visualisation begins
• Initial profiling identifies row counts, fields, ranges and null values
• Frequency tables reveal unexpected categories, codes and inconsistent labels
• Source metadata recorded for ownership, extraction date and refresh cycle

Participants can identify, quantify and document material data quality issues.

• Completeness rates calculated by field, period, source and reporting unit
• Duplicate records detected using candidate keys and exact-match checks
• Validity rules test permitted ranges, formats and reference values
• Consistency checks compare related fields, totals and status combinations
• Outlier screening separates plausible extremes from recording or extraction errors
• Data quality logs document issue severity, treatment and residual risk

Participants can apply transparent cleaning rules and create analysis-ready variables.

• Missing values classified as structural, random, unknown or not applicable
• Imputation choices compared with deletion, flagging and sensitivity analysis
• Text fields cleaned using strip, replace, split and regular expressions
• Dates parsed, validated and converted into reporting periods and durations
• Categorical values recoded through mapping tables and controlled labels
• Derived variables calculated from raw measures using stated business rules

Participants can join related tables and reshape data for analysis and reporting.

• Primary and foreign keys assessed before merging transactional datasets
• Inner, left, right and outer joins selected for stated purposes
• Merge validation detects one-to-one, one-to-many and many-to-many relationships
• Concatenation combines period files while preserving source and version fields
• Pivot tables summarise measures by categories, periods and reporting units
• Wide and long formats converted using melt and pivot operations

Participants can summarise distributions and select measures appropriate to the data.

• Mean, median and mode compared for symmetric and skewed distributions
• Range, variance, standard deviation and interquartile range quantify dispersion
• Quantiles identify distribution thresholds for service, risk and performance measures
• Grouped summaries calculate counts, rates and aggregates by segment
• Skewness and kurtosis considered when interpreting numerical distributions
• Distribution tables distinguish denominator changes from genuine performance movements

Participants can calculate traceable performance measures from detailed records.

• Numerators, denominators and inclusion rules specified before metric calculation
• Rates, ratios and percentages calculated with zero-denominator safeguards
• Weighted averages applied where volumes or exposure differ materially
• Period-over-period change separated into absolute and percentage movements
• Cohort measures track retention, completion and repeat-use patterns
• Metric definitions documented with owner, source, formula and refresh frequency

Participants can assess sampling approaches and express uncertainty in estimates.

• Population, sample frame and sampling unit defined for analysis
• Simple random, systematic, stratified and cluster sampling compared
• Sampling bias identified through coverage, nonresponse and selection mechanisms
• Standard error distinguishes estimate precision from observed data variation
• Confidence intervals calculated and interpreted for means and proportions
• Sample-size reasoning considers effect size, power and practical constraints

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

• Null and alternative hypotheses expressed in measurable, testable terms
• Type I and Type II errors linked to decision consequences
• P-values interpreted alongside confidence intervals and practical significance
• Independent and paired t-tests compare means under stated assumptions
• Chi-square tests assess association between categorical variables and outcomes
• Nonparametric Mann-Whitney tests address ordinal or non-normal distributions

Participants can quantify association without making unsupported causal claims.

• Scatterplots inspect direction, form, spread and influential observations first
• Pearson correlation measures linear association between continuous variables
• Spearman rank correlation assesses monotonic relationships with ordinal data
• Correlation matrices screened for redundant predictors and suspicious relationships
• Confounding variables considered before attributing observed associations to causes
• Association findings reported with coefficients, uncertainty and stated limitations

Participants can build, assess and explain multiple linear regression models.

• Regression questions define outcome, predictors, population and decision use
• Ordinary least squares estimates coefficients for continuous outcome variables
• Categorical predictors encoded using reference categories and dummy variables
• Model outputs interpreted through coefficients, confidence intervals and adjusted R-squared
• Residual plots assess linearity, homoscedasticity and influential cases
• Multicollinearity checked using correlation patterns and variance inflation factors

Participants can model binary outcomes and evaluate classification performance.

• Binary outcomes coded consistently with defined positive-event interpretation
• Logistic regression estimates event probabilities from predictor variables
• Odds ratios translated into practical statements with confidence intervals
• Classification thresholds selected according to operational error costs
• Confusion matrices report true positives, false positives and missed cases
• Precision, recall and ROC curves compare binary classification performance

Participants can prepare time-series data and identify meaningful temporal patterns.

• Time indexes constructed with consistent frequency, timezone and period boundaries
• Trend, seasonality and irregular variation separated through visual inspection
• Rolling averages smooth short-term variation while retaining recent movement
• Period comparisons use aligned days, months and equivalent reporting windows
• Lagged variables test delayed relationships between operational measures
• Forecast baselines use naïve, seasonal naïve and moving-average methods

Participants can select chart forms and design choices that support accurate interpretation.

• Audience, decision and analytical question determine chart selection criteria
• Bar, line, scatter and distribution charts matched to data relationships
• Pre-attentive attributes direct attention through position, length and colour
• Axis scales, zero baselines and intervals chosen without misleading comparisons
• Titles and annotations state findings rather than merely name measures
• Colour palettes distinguish categories while remaining accessible to colour-vision differences

Participants can produce publication-ready statistical charts with Python code.

• Matplotlib figure and axes objects control chart structure precisely
• Histograms, boxplots and density plots reveal distributional characteristics clearly
• Error bars display confidence intervals and estimate uncertainty explicitly
• Faceted charts compare segments without obscuring scale or sample differences
• Annotations highlight threshold breaches, turning points and material outliers
• Export settings produce suitable PNG, SVG and PDF reporting graphics

Participants can create focused interactive charts for exploratory and management reporting.

• Plotly Express creates interactive charts from tidy Pandas DataFrames
• Hover labels expose exact values without overcrowding static chart areas
• Filters and dropdown controls support focused exploration of reporting segments
• Linked chart views compare time, geography, category and performance measures
• Interactive outputs tested for readable labels, units and default states
• HTML exports share standalone visual reports without notebook dependencies

Participants can structure analytical work for rerunning, review and controlled handover.

• Notebook sections separate ingestion, cleaning, analysis, charts and conclusions
• Parameters centralise reporting dates, thresholds, file paths and assumptions
• Assertions stop processing when expected data conditions are not met
• Reusable functions reduce repeated code and inconsistent calculation logic
• Git commits record meaningful analytical changes and review points
• README files explain sources, dependencies, execution order and known limitations

Participants can apply practical controls for sensitive data and reviewed analytical outputs.

• Personally identifiable information identified before data extraction and sharing
• Data minimisation limits fields to stated analytical requirements only
• Access controls distinguish raw, derived and published dataset permissions
• Aggregation and suppression reduce disclosure risk in small reporting groups
• Peer review checks formulas, assumptions, code logic and chart claims
• Analytical audit trails link source records, transformations, outputs and decisions

Participants can build, review and present a complete Python analysis from source data to evidence-based recommendations.

• Analytical brief defines decision question, stakeholders, measures and constraints
• Source datasets profiled, cleaned and combined using documented transformation rules
• Statistical methods selected according to variable type and analytical question
• Regression, comparison or time-series findings tested against stated assumptions
• Visual report presents key findings, uncertainty, limitations and recommendations
• Peer review challenges code, calculations, interpretation and presentation narrative