SPSS for Health Research Training Course

Participants can structure health datasets around appropriate units of analysis and documented source fields.

• Identifying patient, encounter, household and facility units of analysis
• Wide and long formats for repeated clinical measurements
• Creating unique identifiers while protecting direct personal identifiers
• Importing codebooks and mapping source fields to SPSS variables
• Restructuring repeated measures with VARSTOCASES and CASESTOVARS procedures
• Documenting provenance, transformations and ownership in a data dictionary

Participants can identify, correct and document data-quality issues before statistical analysis.

• Frequencies and descriptive tables reveal impossible values and coding errors
• Range checks, logical checks and cross-field consistency rules
• Duplicate-record detection using identifiers, dates and selected clinical fields
• Classifying missingness as MCAR, MAR or MNAR before analysis
• Recoding special missing-value codes without overwriting source values
• Producing a data-cleaning log with query resolutions and rerun syntax

Participants can create documented analysis variables, scores, dates and aggregated health indicators.

• RECODE and COMPUTE commands create analysis-ready categorical variables
• Date functions calculate age, follow-up time and service intervals
• Deriving BMI, blood-pressure categories and clinical threshold indicators
• Aggregating patient records to household, facility and district measures
• Selecting cases and splitting files without corrupting the master dataset
• Creating labelled composite scores from validated questionnaire items

Participants can summarise health data accurately and select clear displays for different outcome types.

• Measurement scales determine appropriate summaries, tables and graphical displays
• Means, medians, proportions and rates reported with suitable denominators
• Explore procedure assesses distribution shape, outliers and normality diagnostics
• Confidence intervals for means and proportions using SPSS output
• Stratified descriptive tables compare outcomes by sex, age and location
• Histograms, boxplots and bar charts follow clear public-health reporting conventions

Participants can calculate and interpret core epidemiological measures for population and service data.

• Prevalence, incidence proportion and incidence rate calculated from study data
• Risk ratio, odds ratio and rate ratio interpreted correctly
• Person-time denominators constructed for cohort and surveillance datasets
• Direct age standardisation compares populations with different age structures
• Crude and stratified estimates expose potential confounding patterns
• Presenting effect estimates with 95% confidence intervals and denominators

Participants can account for sampling weights, strata and clusters in survey analysis.

• Probability and non-probability sampling designs shape valid population inferences
• Sampling weights, strata and clusters specified in Complex Samples plans
• Design effects and intracluster correlation affect precision estimates
• Finite population correction and weighting choices documented transparently
• Survey frequencies and cross-tabulations produced with design-based standard errors
• Interpreting weighted estimates without confusing sample counts and population totals

Participants can select, run and interpret appropriate bivariate tests for common health research questions.

• Research questions translated into null hypotheses, estimands and test choices
• Independent and paired t-tests analyse continuous health outcomes
• Chi-square, Fisher's exact and McNemar tests compare categorical outcomes
• One-way ANOVA and post-hoc comparisons examine multiple group means
• Non-parametric Mann-Whitney, Wilcoxon and Kruskal-Wallis alternatives selected appropriately
• Effect sizes, confidence intervals and p-values reported together

Participants can assess associations and fit simple linear regression models for continuous outcomes.

• Pearson and Spearman correlations assess linear and monotonic associations
• Scatterplots identify nonlinearity, clusters and influential observations before modelling
• Simple linear regression estimates associations between exposure and outcome
• Residual plots assess linearity, homoscedasticity, normality and influential cases
• Regression coefficients interpreted in original units with confidence intervals
• Correlation findings distinguished from causal conclusions in health reports

Participants can build and interpret adjusted linear regression models for continuous health outcomes.

• Multiple linear regression adjusts continuous-outcome estimates for confounders
• Confounder selection combines subject-matter knowledge, DAGs and statistical criteria
• Categorical predictors coded with indicator variables and reference groups
• Interaction terms test effect modification by sex, age or setting
• Variance inflation factors diagnose multicollinearity among candidate predictors
• Adjusted coefficients translated into decision-relevant public-health statements

Participants can model binary health outcomes and interpret adjusted odds ratios and predicted probabilities.

• Binary logistic regression models disease status, treatment uptake and mortality
• Odds ratios, adjusted odds ratios and predicted probabilities interpreted
• Reference categories and categorical covariates specified in SPSS dialogs
• Goodness-of-fit assessed using Hosmer-Lemeshow, classification and ROC measures
• Influential cases and sparse cells checked before final inference
• Logistic regression tables prepared to STROBE reporting expectations

Participants can analyse counts and event rates while accounting for exposure time and overdispersion.

• Poisson regression models counts of events with person-time offsets
• Negative binomial models address overdispersion in service-use data
• Rate ratios interpreted for surveillance, admissions and outbreak outcomes
• Zero-inflation considered when structural zeros dominate count distributions
• Model fit compared using deviance, Pearson statistics and information criteria
• Offset variables specified correctly for unequal observation periods

Participants can analyse time-to-event outcomes using Kaplan-Meier and Cox regression procedures.

• Kaplan-Meier curves estimate time to death, recovery or loss-to-follow-up
• Log-rank tests compare survival functions between exposure groups
• Cox proportional hazards regression estimates adjusted hazard ratios
• Proportional hazards assumptions checked using log-minus-log plots
• Censoring, competing events and follow-up definitions documented precisely
• Survival tables and plots prepared for clinical and programme audiences

Participants can analyse change over time using repeated-measures ANOVA and linear mixed models.

• Paired designs distinguish within-person change from between-group differences
• Repeated-measures ANOVA analyses scheduled observations under key assumptions
• Linear mixed models handle unequal visits and incomplete longitudinal records
• Subject identifiers and time variables structured for mixed-model estimation
• Covariance structures compared using information criteria and residual patterns
• Interpreting time-by-group interactions for longitudinal intervention evaluation results

Participants can prepare transparent statistical outputs for research reports, operational decisions and publication.

• STROBE and CONSORT items guide transparent observational and trial reporting
• Publication tables separate descriptive, crude and adjusted estimates clearly
• Chart selection matches outcome type, audience and decision question
• SPSS pivot tables formatted with labels, denominators and confidence intervals
• Syntax, output and data dictionaries packaged for reproducible handover
• Confidentiality checks suppress small cells and remove disclosure risks

Participants can build, review and present a complete reproducible SPSS analysis addressing a health research question.

• Framing a health research question, population, exposure and outcome
• Preparing an analysis dataset through documented cleaning and derived variables
• Selecting descriptive, bivariate and multivariable methods from study design
• Running SPSS syntax and verifying output against prespecified analytic decisions
• Interpreting estimates, assumptions and limitations for a nontechnical audience
• Presenting a reproducible health analysis with tables, figures and recommendations