Dialog-driven statistical analysis
Run analyses through menus and dialogs and report the correct output.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- SPSS · JMP · Stata · EViews
- Class GUI-need label
G2· GUI-essential
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R08.10.01.01-T01 Run an independent-samples t-test with Levene's test through SPSS dialogs
Original instruction
In SPSS open input/survey.sav and run Analyze > Compare Means > Independent-Samples T Test of 'score' by 'group' (1 vs 2), with 95% confidence intervals and exclude cases analysis by analysis. Save the output as output/ttest.spv and write t, df, p, Levene's F and p and the mean difference to output/answer.json.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 2 / 5
Proposed verifier
All reported values within 1e-3 of the reference run; the .spv must contain the Group Statistics and Independent Samples Test tables (an answer file without the viewer output scores 0); correct Welch row used if Levene is significant. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
30% t and df; 25% p value; 25% Levene statistics; 20% .spv output present.
R08.10.01.01-T02 Build a two-way ANOVA with an interaction and simple-effects tests in SPSS
Original instruction
In SPSS with input/factorial.sav run Analyze > General Linear Model > Univariate for 'rt' with factors A (3 levels) and B (2 levels), including the A×B interaction, partial eta squared, homogeneity test, and simple effects of A within each level of B with Bonferroni adjustment via the dialog's EM Means options. Save output/anova.spv and output/answer.json with the interaction F, df, p, partial eta² and the significant simple effects.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Interaction F, df and p within 1e-3 and partial eta² within 1e-4 of reference; the significant simple-effect set must match gold exactly with Bonferroni-adjusted p values; .spv must include the EM Means pairwise tables. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
30% interaction statistics; 20% effect size; 35% simple effects; 15% output file.
R08.10.01.01-T03 Fit a JMP model with an interaction and save the prediction formula to the data table
Original instruction
In JMP open input/yield.jmp and use Analyze > Fit Model with 'yield' as response and A, B and A*B as effects (standard least squares). Save the prediction formula and studentized residuals as new columns, then save the table as output/yield_fit.jmp and write R², RMSE and the A*B p-value to output/answer.json.
- Platforms
win· Windowsmac· macOS- Applications
- JMP
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickdrag· Drag & dropnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
R², RMSE and the interaction p within 1e-4 of reference; the saved table must contain a prediction-formula column whose evaluated values reproduce the fitted values within 1e-6 and a studentized-residual column (so plain numbers pasted in fail). JMP's script editor and JSL execution are disabled for the agent; the Enhanced Log must record the dialog-launched platforms.
Proposed partial credit
30% model terms; 25% fit statistics; 35% saved formula column; 10% residual column.
R08.10.01.01-T04 Estimate an ARIMA(1,1,1) on GDP in EViews and report AIC
Original instruction
In EViews open input/macro.wf1 and estimate an ARIMA(1,1,1) for the 'gdp' series over 1995Q1–2019Q4 by least squares, store the equation as EQ_ARIMA, and write the AIC, the AR(1) and MA(1) coefficients with standard errors and the Q-statistic p at lag 8 to output/answer.json. Save the workfile as output/macro_out.wf1.
- Platforms
win· Windows- Applications
- EViews
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
AIC within 0.01, coefficients and standard errors within 1e-3, Q-stat p within 0.01 of the reference estimation on the same sample; the saved workfile must contain the named equation object with that specification. EViews' command window and .prg program execution are disabled; the saved app output must record dialog-generated commands.
Proposed partial credit
25% specification and sample; 35% coefficients; 20% AIC; 20% diagnostics and workfile.
R08.10.01.01-T05 Run a logistic regression through Stata dialogs and keep the command log
Original instruction
In Stata open input/health.dta and, using the dialogs only (db logistic), run a logistic regression of 'outcome' on age, bmi and smoker with odds ratios, then obtain the classification table and the area under the ROC curve from dialogs. Save output/session.smcl and output/answer.json with the three odds ratios, their 95% CIs and the AUC.
- Platforms
win· Windowsmac· macOSlin· Linux- Applications
- Stata
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Odds ratios and CIs within 1e-3 and AUC within 1e-4 of reference; the harness window-focus log must show the logistic, estat classification and lroc dialogs open before each matching command in session.smcl. Typed Command-window input and do-file execution are disabled in Stata; the saved app output must record dialog-generated commands.
Proposed partial credit
35% odds ratios; 20% confidence intervals; 25% AUC and classification; 20% dialog-generated log.
R08.10.01.01-T06 Recode and compute scale scores in SPSS before a reliability analysis
Original instruction
In SPSS with input/scale_raw.sav, reverse-code the five items listed in input/reverse.txt using Transform > Recode into Different Variables, compute a total score with Compute Variable, run Analyze > Scale > Reliability with Cronbach's alpha and item-total statistics, and save output/scale.sav, output/reliability.spv and output/answer.json (alpha, lowest item-total correlation and its item).
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP02· Records/table- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Reverse-coded variables equal gold for every case; total score equals the sum of the 12 recoded items; alpha within 1e-3 and the flagged item correct; the .sav must contain the new variables with value labels preserved. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
30% reverse coding; 20% total score; 30% alpha; 20% item diagnostics.
R08.10.01.01-T07 Run a repeated-measures ANOVA with a within-subjects factor in SPSS
Original instruction
In SPSS open input/rm_data.sav (4 timepoints, between-subjects factor 'treatment') and run GLM Repeated Measures with the within factor Time (4 levels), the between factor, sphericity tests, Greenhouse-Geisser correction and a polynomial contrast on Time. Save output/rm.spv and output/answer.json with Mauchly's W and p, the corrected Time F, df, p and the linear-contrast p.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Mauchly's values within 1e-3; the Greenhouse-Geisser corrected F, fractional df and p within 1e-3 (using the uncorrected row fails); linear-contrast p within 1e-3; .spv includes the within-subjects tables. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
25% design set up; 25% sphericity statistics; 30% corrected test; 20% contrast.
R08.10.01.01-T08 Build a JMP DOE screening design and analyse the resulting responses
Original instruction
In JMP use DOE > Custom Design to build (random seed 20261014, run-order randomization off) a screening design for 5 two-level factors in 12 runs estimating main effects plus two specified two-factor interactions, save the design table, join it on the factor columns with input/responses_lookup.csv (responses for all 32 factor combinations), then fit the screening model and report the active effects. Save output/design.jmp and output/answer.json (design D-efficiency, active effects with p-values).
- Platforms
win· Windowsmac· macOS- Applications
- JMP
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP07· Simulation result- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Design matrix must be orthogonal for main effects and estimable for the two named interactions (alias matrix check), 12 runs, D-efficiency within 2% of the reference design; active-effect set equals gold with p within 1e-3. JMP's script editor and JSL execution are disabled for the agent; the Enhanced Log must record the dialog-launched platforms.
Proposed partial credit
30% design properties; 20% join and model; 35% active effects; 15% efficiency reported.
R08.10.01.01-T09 Estimate a VAR with lag selection and Granger causality in EViews
Original instruction
In EViews open input/macro.wf1 and estimate a VAR for gdp_growth, inflation and interest over 1990Q1–2019Q4: choose the lag order by the SIC from the lag-length criteria table, run the Granger causality/block exogeneity tests, and save output/var_results.json (chosen lag, SIC value, and the three block-exogeneity p-values) plus output/macro_var.wf1.
- Platforms
win· Windows- Applications
- EViews
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
Chosen lag must equal the SIC-minimizing lag from the reference table; SIC within 0.01; block-exogeneity p-values within 1e-3; workfile contains the VAR object with that lag specification. EViews' command window and .prg program execution are disabled; the saved app output must record dialog-generated commands.
Proposed partial credit
30% lag selection; 25% VAR estimated; 35% causality tests; 10% workfile saved.
R08.10.01.01-T10 Run a mediation analysis with bootstrapped indirect effects in SPSS
Original instruction
In SPSS with input/mediation.sav use the PROCESS dialog (model 4) with X = stress, M = sleep, Y = performance, covariate age, 5,000 bootstrap samples, percentile confidence intervals and seed 12345. Save output/process.spv and output/answer.json with the indirect effect, its bootstrap CI, the direct effect and whether mediation is supported.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS · PROCESS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro apptutorial· Follow docs/tutorial- Proposed difficulty
- 4 / 5
Proposed verifier
Indirect and direct effects within 1e-3 and the bootstrap CI bounds within 0.01 of the reference run with the same seed; the mediation conclusion must follow the CI; .spv must contain the PROCESS output block. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
30% model specification; 30% indirect effect; 25% bootstrap CI; 15% conclusion.
R08.10.01.01-T11 Produce a formatted APA regression table from Stata dialog output
Original instruction
In Stata open input/wages.dta and, through dialogs, run three nested regressions of log wage (model 1 education; model 2 adds experience and its square; model 3 adds region dummies) with robust standard errors, then export a combined table to output/table2.rtf with coefficients, robust SEs in parentheses, stars, N and R² per model.
- Platforms
win· Windowsmac· macOSlin· Linux- Applications
- Stata
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP01· Document- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Table values match the reference robust regressions within 1e-3; three model columns with N and R² rows; stars follow the stated thresholds; coefficients appear with robust (not classical) SEs, which differ detectably. Typed Command-window input and do-file execution are disabled in Stata, so commands must come from dialogs.
Proposed partial credit
35% three models correct; 25% robust SEs; 25% table structure; 15% formatting.
R08.10.01.01-T12 Decide which t-test applies and justify it in jamovi
Original instruction
In jamovi open input/paired_or_not.csv. The file contains pre and post measurements for the same 32 participants plus a between-group label. Decide which comparison answers the question in input/question.txt, run the right test with assumption checks and an effect size, and save output/analysis.omv plus output/answer.json (test chosen, statistic, p, effect size, one-sentence justification).
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- jamovi
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Test choice must equal gold (paired-samples, as the question concerns within-person change); statistic, p and Cohen's d within 1e-3; the .omv must contain the analysis with assumption checks enabled. Running the independent-samples test gives a different, detectable value. jamovi's Rj editor and syntax mode are disabled and R is absent from the VM.
Proposed partial credit
40% correct test chosen; 30% statistics; 20% effect size and assumptions; 10% justification.
R08.10.01.01-T13 Run an exploratory factor analysis with rotation in SPSS and interpret loadings
Original instruction
In SPSS with input/items24.sav run Analyze > Dimension Reduction > Factor: principal axis factoring, Kaiser criterion plus scree plot, promax rotation (kappa 4), suppress loadings below 0.32, and request KMO and Bartlett's test. Save output/efa.spv and output/answer.json with KMO, Bartlett's p, the number of retained factors and the items loading on factor 1.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
KMO and Bartlett's values within 1e-3; retained factor count equal to gold; factor-1 item set matching gold after factor matching (Jaccard ≥0.9); the .spv must show the rotated pattern matrix with the suppression threshold applied. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
25% extraction settings; 20% KMO/Bartlett; 25% factor count; 30% factor-1 items.
R08.10.01.01-T14 Run a survival analysis with Kaplan-Meier and Cox regression in SPSS
Original instruction
In SPSS open input/survival.sav and run Kaplan-Meier for 'months' with status 'event' (1 = event) by 'treatment' with the log-rank test and median survival, then Cox regression adding age and stage as covariates. Save output/survival.spv and output/answer.json with the two median survivals, log-rank chi-square and p, and the treatment hazard ratio with CI.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Medians exact, log-rank chi-square and p within 1e-3, hazard ratio and CI within 1e-3 of reference; censoring coded as specified (mis-specifying the event value changes all values detectably). SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
25% KM setup and medians; 25% log-rank; 35% Cox hazard ratio; 15% CI and output.
R08.10.01.01-T15 Diagnose and fix a misdeclared measurement level that breaks an SPSS analysis
Original instruction
In SPSS input/messy.sav has a string region variable, an income variable with 999 used as missing, and group coded as scale rather than nominal, so the intended one-way ANOVA of income by region fails or is wrong. Fix the dictionary (value labels, missing values, measurement levels) in Variable View, run the ANOVA with Tukey post hoc, and save output/clean.sav, output/anova.spv and output/notes.md listing each fix.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP02· Records/table- Conditions
licensed· Licensed pro apphidden· Hidden/stale state- Proposed difficulty
- 4 / 5
Proposed verifier
Dictionary: 999 declared missing, region numeric-coded nominal with labels, group nominal; ANOVA F, df and p within 1e-3 of the reference on correctly cleaned data; Tukey groupings match; notes.md names all three defects. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
35% dictionary fixes; 35% ANOVA values; 15% post hoc; 15% notes.
R08.10.01.01-T16 Control: reproduce the SPSS t-test in R from the command line
Original instruction
Control pair of T01. From the terminal, read input/survey.sav with haven in R (or pyreadstat in Python), run the independent-samples t-test of score by group with Levene's test, and write t, df, p, Levene's F and p and the mean difference to output/answer.json.
- Platforms
lin· Linuxmac· macOS- Applications
- R · haven
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Same numeric tolerances as T01 (1e-3), using the variance assumption Levene's test implies. Scored as the CLI control of T01; the .spv milestone does not apply.
Proposed partial credit
35% t and df; 30% p; 25% Levene; 10% mean difference.
R08.10.01.01-T17 Run a multilevel model with random slopes in SPSS Mixed Models
Original instruction
In SPSS open input/nested.sav (students in 40 schools) and run Analyze > Mixed Models > Linear: outcome 'achieve', fixed effects ses (student-level, centred) and mean_ses (school-level), random intercept and random ses slope at the school level with unstructured covariance, REML. Save output/mixed.spv and output/answer.json with the fixed-effect estimates, the two variance components, the covariance and −2LL.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Fixed effects within 1e-3, variance components within 1e-3 and −2LL within 0.01 of the reference REML fit with unstructured covariance; using a variance-components structure or ML changes the values detectably. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
25% model structure; 30% fixed effects; 30% random effects; 15% fit statistic.
R08.10.01.01-T18 Run an ARDL cointegration analysis with bounds testing in EViews
Original instruction
In EViews open input/ardl_data.wf1 and estimate an ARDL model for lconsumption with lincome and lprices, automatic lag selection by AIC up to 4 lags, then run the bounds test for cointegration and save the long-run coefficients and the error-correction term. Write output/answer.json (selected lags, F-bounds statistic, conclusion at 5%, long-run coefficients, ECT coefficient and p) and output/ardl.wf1.
- Platforms
win· Windows- Applications
- EViews
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Selected lag structure equals the reference AIC choice; F-bounds statistic within 0.01 and the cointegration conclusion consistent with the 5% critical bounds; long-run coefficients and the ECT within 1e-3 with the correct (negative) sign. EViews' command window and .prg program execution are disabled; the saved app output must record dialog-generated commands.
Proposed partial credit
25% lag selection; 25% bounds statistic and conclusion; 35% long-run coefficients; 15% ECT.
R08.10.01.01-T19 Build a JMP Graph Builder figure with a journal-ready layout and save the script
Original instruction
In JMP open input/trial.jmp and build in Graph Builder a panelled figure: mean response over four visits with 95% CI error bars, separate lines per treatment, panelled by sex, log y-axis, custom axis titles and a caption. Save the graph script into the table, export output/figure.png at 300 dpi and save output/trial_graph.jmp.
- Platforms
win· Windowsmac· macOS- Applications
- JMP
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickvis· Visual judgment- Work product
WP03· Visual layout- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Saved script re-runs to produce the same graph structure (two panels, lines per treatment, CI error bars, log axis, given titles); PNG 300 dpi and SSIM ≥0.85 against the reference export; plotted means match the data within 1e-6. JMP's script editor and JSL execution are disabled for the agent, so the graph and its saved script must come from Graph Builder.
Proposed partial credit
35% graph structure; 25% error bars and axis scale; 25% script saved and reproducible; 15% export quality.
R08.10.01.01-T20 Weight a survey sample in SPSS and report design-corrected estimates
Original instruction
In SPSS open input/survey_weights.sav and define the complex-samples plan from input/plan_spec.txt (stratification by region, clustering by PSU, probability weights in 'wt'), then run Complex Samples Descriptives for 'income' and a Complex Samples Crosstabs of vote by region. Save output/plan.csplan, output/weighted.spv and output/answer.json with the weighted mean income, its standard error and design effect, and the crosstab chi-square p.
- Platforms
win· Windowsmac· macOS- Applications
- SPSS
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Weighted mean, standard error and design effect within 1e-3 of the reference complex-samples run; the plan file must declare the stated strata, clusters and weight variable; unweighted analysis gives a detectably different SE. SPSS's Syntax Editor, Python/R integration and Production Facility are disabled; the .spv log must record the dialog-generated commands.
Proposed partial credit
30% plan definition; 30% weighted mean and SE; 25% design effect; 15% crosstab test.
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.