Your thesis needs an ANOVA, and the university's SPSS licence only works in the computer room. The usual online answer is "just install JASP". That is half right. Common analyses are covered, but what you really need when you switch software and the numbers do not match is to know where the difference comes from.
Bottom line: to run routine analyses quickly from menus, start with jamovi or JASP. For long-term research and fully reproducible workflows, learn R. None of the three is a full replacement for SPSS, so check carefully if you need particular modules or your institution requires it.
What was tested (2026-11-09): the R code was actually run on R 4.3.3 with car 3.1-2. The code was run in the editorial environment (Linux). JASP and jamovi were not operated; section 4 gives the R reference values only. All numbers here come from R only.
1. What each tool is
| | JASP | jamovi | R | |---|---| | How you use it | Menus, results update live | Spreadsheet plus menus, results update live | Write code | | Cost | Free | Desktop free; a cloud version also exists | Free | | Licence | AGPL-3.0 (repository COPYING) | AGPL-3.0 (repository LICENSE.md; some components differ) | GPL-2.0 (ZIRULAB tool record) | | Character | Frequentist and Bayesian methods side by side | Spreadsheet-like, extensible with modules | Widest range of methods, fully scriptable | | Learning curve | Low | Low | Higher |
JASP and jamovi both build on the R ecosystem, but that does not mean their defaults match R exactly.
2. One data set, one task
This article uses PlantGrowth, which ships with R: plant dry weight in 3 groups of 10. It is public and small, and can be exported to CSV for JASP or jamovi.
Research question: do the three groups differ in mean dry weight?
3. The full workflow in R
You need R and the car package (install.packages("car")). This code runs as it stands and also exports plantgrowth.csv.
library(car)
d <- PlantGrowth
write.csv(d, "plantgrowth.csv", row.names = FALSE)
# 1. Descriptive statistics
aggregate(weight ~ group, d, function(x) c(n = length(x), mean = mean(x), sd = sd(x)))
# 2. Assumptions
by(d$weight, d$group, shapiro.test)
leveneTest(weight ~ group, d) # centred on the median by default
# 3. One-way ANOVA
fit <- aov(weight ~ group, d)
summary(fit)
# 4. Post hoc comparisons
TukeyHSD(fit)
# 5. Without equal variances / non-parametric
oneway.test(weight ~ group, d)
kruskal.test(weight ~ group, d)
# 6. Record versions for your methods section
sessionInfo()
Actual output:
| Item | Result in R 4.3.3 (reference value) |
|---|---|
| Means of ctrl / trt1 / trt2 | 5.032 / 4.661 / 5.526 |
| SDs | 0.583 / 0.794 / 0.443 |
| Levene (median-centred) | F(2, 27) = 1.119, p = 0.341 |
| One-way ANOVA | F(2, 27) = 4.846, p = 0.016, η² = 0.264 |
| Welch ANOVA | F(2, 17.13) = 5.181, p = 0.017 |
| Kruskal-Wallis | χ²(2) = 7.988, p = 0.018 |
| Tukey: trt2 − trt1 | difference 0.865, adjusted p = 0.012 |
| Tukey: trt1 − ctrl | difference −0.371, adjusted p = 0.391 |
| Tukey: trt2 − ctrl | difference 0.494, adjusted p = 0.198 |
In words: the groups differ overall, but only the difference between trt2 and trt1 reaches the 0.05 level after Tukey correction.
4. Checking in JASP and jamovi
Import plantgrowth.csv, set weight as the dependent variable and group as the grouping variable, run a one-way ANOVA, and tick the homogeneity test and Tukey post hoc comparisons. Menu names may vary by version.
| Item | Result in R 4.3.3 (reference value) |
|---|---|
| ANOVA F | 4.846 |
| ANOVA p | 0.016 |
| Tukey trt2 − trt1 p | 0.012 |
This analysis was not run in JASP or jamovi for this article, so no numbers are given for them. The table shows the R results, which you can use as reference values: run the same analysis in your version, and if F, p or the Tukey result differs, check the settings in the order given in section 5. Script and data: r_stats_demo.R and plantgrowth.csv; the full output is in r_stats_output.txt.
5. When results disagree, check in this order
All numbers below were actually produced in R.
1. Student or Welch t-test. For ctrl versus trt1, t is 1.1913 either way, but df and p differ: Student has df = 18 and p = 0.2490; Welch has df = 16.52 and p = 0.2504. R's t.test() defaults to Welch, and other software may default or display differently.
2. Levene centred on the mean or median. On the same data, the median gives F = 1.119, p = 0.341 and the mean gives F = 1.237, p = 0.306.
3. Type of sums of squares. This is the usual cause of large disagreements. Simulated unbalanced two-way data (48 cases; cells of 22, 8, 6 and 12) for factor A:
| Type | SS for A | F | p | |---|---| | Type I (A entered first) | 56.437 | 29.32 | 2.4 × 10⁻⁶ | | Type II | 19.295 | 10.03 | 0.0028 | | Type III | 19.4 | 10.10 | 0.0027 |
With unequal group sizes, the three mean different things and give different results. aov() and anova() use Type I, which also depends on variable order; SPSS GLM generally defaults to Type III. Type III also needs sum-to-zero contrasts (contr.sum), otherwise it does not mean what you intend.
4. Missing values. R's mean() returns NA by default when values are missing and calculates with na.rm = TRUE. Software and analyses may differ in listwise or pairwise handling.
5. Quantile definitions. For the same ctrl data, the first quartile is 4.53, 4.5225 and 4.55 with R's types 1, 6 and 7. Small differences in box plots and descriptives often come from this.
6. Variable type and coding. A numerically coded group treated as continuous gives an entirely different model.
7. One or two tails, correction method, rounding. Compare each before reporting.
6. What cannot simply be replaced in SPSS
IBM's product pages list additional capabilities in SPSS Statistics such as Custom Tables, Complex Samples, Exact Tests and Forecasting, some of them in separate modules or higher editions. Common practical differences are:
- Complex sampling and custom tables: these are specialised features listed by IBM. Before switching, confirm an alternative suits your design.
- Syntax: existing SPSS Syntax does not run directly elsewhere.
- Institutional and journal requirements: some procedures specify SPSS. That is a compliance question, not a statistical one.
- Support: a commercial contract includes support; open-source software relies mainly on its community.
Prices and licensing vary by region and institution, so check IBM's official quotation. Also ask whether your university has a campus licence. Sometimes the easiest option is the one you already have.
7. Which to choose
| Reader | Suggestion |
|---|---|
| Undergraduate, coursework or thesis | Start with jamovi or JASP; if your teacher specifies SPSS, follow that |
| Graduate student planning long-term research | Start with jamovi or JASP while learning R, because supervisors and reviewers may ask for reproduction |
| Researcher with recurring data and analyses | R as the main tool, saving scripts and sessionInfo() for key analyses |
| Dependent on SPSS complex sampling, custom tables or existing syntax | Keep SPSS for now and run a small migration test first |
Whichever you choose, state the software name, version, test type and correction method in your methods.