Chapter 16. SPSS, R, Stata, SAS, Jamovi, JASP, Excel and Manual Checks
16.1 Software does not replace statistical reasoning
Software executes procedures; the researcher remains responsible for data coding, model selection, assumptions, interpretation and reporting. Use syntax or scripts rather than relying exclusively on point-and-click menus. A reproducible workflow makes it possible to rerun analyses after corrections and to audit exactly how results were produced.
| Tool | Typical strengths | Cautions |
|---|---|---|
| IBM SPSS Statistics | Accessible menus, common in social/business sciences, broad procedures | Commercial licensing; syntax should be saved |
| R | Free, extensive statistical ecosystem, reproducible scripts and graphics | Learning curve; package versions and code should be documented |
| Stata | Strong applied statistics, econometrics, panel and survey workflows | Commercial licensing; preserve do-files and logs |
| SAS | Enterprise analytics, regulated and large-scale workflows | Commercial ecosystem; document procedures |
| jamovi | Accessible interface with reproducible analysis options | Check module capabilities and model assumptions |
| JASP | Accessible interface, classical and Bayesian methods | Check procedure details and version |
| Excel | Data review, simple summaries, transparent calculations | Risk of silent formula errors; limited robust workflow for complex analyses |
| Python | Data engineering, modelling and machine learning | Requires disciplined coding, validation and environment documentation |
| Manual calculation | Useful for understanding formulas and checking simple results | Prone to arithmetic error and impractical for complex models |
16.2 SPSS project workflow
- Create a clearly named .sav file and preserve the original export separately.
- In Variable View, define name, type, label, value labels, missing values and measurement level.
- Run Frequencies for categorical variables and Explore/Descriptives for numeric variables.
- Use Data > Identify Duplicate Cases or appropriate syntax; document decisions.
- Use Transform > Recode into Different Variables for reverse coding; retain original items.
- Compute composite scores only after checking instrument scoring rules and missingness.
- Use Analyze menus to run the prespecified test; save syntax through Paste or a syntax window.
- Inspect diagnostic plots and relevant assumptions, not only the output table.
- Export tables carefully and verify every number against the original output.
- Save syntax, output (.spv), data dictionary and an analysis log.
16.3 Example SPSS syntax
Illustrative syntax only; replace variable names and values with those in the approved dataset. Verify the command options in the installed SPSS version.
FREQUENCIES VARIABLES=gender work_mode. DESCRIPTIVES VARIABLES=engagement_total support_total /STATISTICS=MEAN STDDEV MIN MAX. EXAMINE VARIABLES=engagement_total support_total /PLOT=HISTOGRAM NPPLOT /STATISTICS=DESCRIPTIVES /MISSING=LISTWISE. CORRELATIONS /VARIABLES=support_total engagement_total /PRINT=TWOTAIL SIG /MISSING=PAIRWISE. REGRESSION /DEPENDENT engagement_total /METHOD=ENTER support_total tenure_years work_mode /STATISTICS COEFF OUTS R ANOVA COLLIN TOL /PLOT=SCATTER(*ZRESID,*ZPRED) /RESIDUALS HISTOGRAM(ZRESID) NORMPROB(ZRESID).
16.4 Equivalent R illustration
# Illustrative only; verify names, missingness, coding and assumptions. summary(dat$engagement_total) table(dat$work_mode, useNA = "ifany") cor.test(dat$support_total, dat$engagement_total, use = "complete.obs", method = "pearson") fit <- lm(engagement_total ~ support_total + tenure_years + work_mode, data = dat) summary(fit) confint(fit) par(mfrow = c(2, 2)) plot(fit)
16.5 Manual checks
Manual checks are valuable for simple calculations and verification. For a mean, sum valid observations and divide by the number of valid observations. For a sample variance, sum squared deviations from the sample mean and divide by n−1; the sample standard deviation is the square root. For a simple two-group difference, calculate group means, the difference, a suitable standard error and a confidence interval using the method appropriate to the design. Do not use manual calculation as a substitute for a correct model when data are clustered, weighted, longitudinal or otherwise complex.
16.6 Reproducibility checklist
- Raw data are retained unchanged and access-controlled.
- Every derived variable has a documented rule.
- Software name, version, packages/modules and operating environment are recorded.
- Analysis syntax or scripts run from beginning to end without undocumented manual edits.
- Random seeds are set for stochastic procedures where appropriate.
- Outputs are linked to the reported tables and figures.
- Corrections and exclusions are documented.
- An appropriately de-identified analysis dataset and code can be shared when consent, law, contracts and policy permit.