2. A complete analysis case:
collaborative-regulation sequences
This vignette runs one dataset all the way through and reads
the numbers at each step – not just what to call but
what the output means and what not to over-read. The
data are the bundled group_regulation_long event log:
students’ collaborative regulation-of-learning actions
(plan, monitor, consensus,
discuss, …), one row per action, with a High /
Low achievement label per student.
The arc that emerges, stated up front so the sections connect:
regulation talk has short memory – the immediately
preceding action carries most of the predictive signal – a handful of
two-action routines reproducibly add to it, and high and low
achievers regulate differently, which the permutation test
confirms.
1. The data
data(group_regulation_long)
nrow(group_regulation_long)
#> [1] 27533
head(group_regulation_long)
#> Actor Achiever Group Course Time Action
#> 1 1 High 1 A 2025-01-01 10:27:07 cohesion
#> 2 1 High 1 A 2025-01-01 10:35:20 consensus
#> 3 1 High 1 A 2025-01-01 10:42:18 discuss
#> 4 1 High 1 A 2025-01-01 10:50:00 synthesis
#> 5 1 High 1 A 2025-01-01 10:52:25 adapt
#> 6 1 High 1 A 2025-01-01 10:57:31 consensus
sort(table(group_regulation_long$Action), decreasing = TRUE)
#>
#> consensus plan discuss emotion coregulate cohesion monitor
#> 6797 6623 4267 3075 2133 1839 1516
#> synthesis adapt
#> 729 554
The nine actions are very unevenly used: consensus and
plan dominate, adapt and
synthesis are rare. That imbalance is the most important
fact about the corpus and it echoes through every result – a model that
just guesses consensus will look deceptively good, so the
interesting question is never “what is the modal next action” but
“which histories overturn that default”.
2. Fit
context_tree() reads the long log directly: name the
unit (actor), the clock (time), and the state
(action); it reshapes into one sequence per session and
fits. Sessions are split where the time gap is large.
tree <- context_tree(group_regulation_long,
actor = "Actor", time = "Time", action = "Action",
max_depth = 3L, min_count = 10L)
tree
#> <transitiontrees> 377 nodes, depth <= 3, 9 states [unpruned]
#> alphabet : adapt, cohesion, consensus, coregulate, discuss, emotion, monitor, plan, synthesis
#> fit on : 2000 sequences, 27533 observations
#> smoothing: floor(ymin=0.001, rule=interpolate) min_count = 10
#> (start) n=27533 -> consensus (0.25)
#> |-- adapt n=509 -> consensus (0.47)
#> | |-- consensus n=27 -> cohesion (0.48)
#> | |-- coregulate n=28 -> consensus (0.50)
#> | | `-- consensus n=21 -> consensus (0.43)
#> | |-- discuss n=259 -> consensus (0.47)
#> | | |-- consensus n=60 -> consensus (0.53)
#> | | |-- coregulate n=37 -> consensus (0.43)
#> | | |-- discuss n=48 -> consensus (0.52)
#> | | |-- emotion n=14 -> consensus (0.50)
#> | | |-- monitor n=25 -> consensus (0.40)
#> | | `-- plan n=26 -> consensus (0.42)
#> | |-- monitor n=16 -> consensus (0.50)
#> | `-- synthesis n=140 -> consensus (0.48)
#> | `-- discuss n=107 -> consensus (0.45)
#> |-- cohesion n=1695 -> consensus (0.50)
#> | |-- adapt n=130 -> consensus (0.54)
#> | | |-- consensus n=13 -> consensus (0.61)
#> | | |-- discuss n=65 -> consensus (0.55)
#> | | `-- synthesis n=32 -> consensus (0.56)
#> | |-- cohesion n=45 -> consensus (0.42)
#> | | `-- emotion n=24 -> consensus (0.46)
#> | |-- consensus n=84 -> consensus (0.51)
#> | | |-- cohesion n=11 -> plan (0.45)
#> | | |-- discuss n=13 -> consensus (0.61)
#> | | |-- emotion n=11 -> consensus (0.45)
#> ... 351 more nodes (use as.data.frame(x) or summary(x))
The banner reports the depth, the node count, the alphabet, and the
sequence/observation totals. The root line is the null
model: the next action given no history. Every deeper
context has to beat that to earn its place.
3. Inspect
summary(tree)
#> <transitiontrees summary> 377 nodes, depth <= 3, 9 states [unpruned]
#>
#> pathway depth count likely_next next_probability divergence
#> (start) 0 27533 consensus 0.2468674 NA
#> consensus 1 6329 plan 0.3957971 0.3340303
#> plan 1 6157 plan 0.3742082 0.2341431
#> discuss 1 3951 consensus 0.3211845 0.5556255
#> emotion 1 2837 cohesion 0.3253437 0.5551452
#> coregulate 1 1970 discuss 0.2736041 0.1808906
#> cohesion 1 1695 consensus 0.4979351 0.3149856
#> monitor 1 1433 discuss 0.3754361 0.2283039
#> synthesis 1 652 consensus 0.4630613 0.8917924
#> adapt 1 509 consensus 0.4741100 0.7892874
#> changes_prediction
#> NA
#> TRUE
#> TRUE
#> FALSE
#> TRUE
#> TRUE
#> FALSE
#> TRUE
#> FALSE
#> FALSE
#> # ... 367 more rows (use as.data.frame(tree) for the full table)
model_fit(tree)
#> logLik df nobs AIC BIC perplexity
#> 1 -45464.76 3016 27533 96961.51 121762.5 5.213661
Perplexity is the readable scalar: the effective number of equally
likely next actions. The uniform baseline is 9 (nine actions, no
knowledge); the fitted tree’s 5.21 says recent history collapses nine
possibilities to about 5.2. Real structure – but this is
in-sample and the tree is over-grown, so read it as an
optimistic bound. Sections 6 and 7 give the honest figure.
4. The pathway tables
Three named verbs each fix a useful sort over the one canonical
schema.
common_pathways(tree, top = 8) # the highways
#> pathway depth count likely_next next_probability divergence
#> 1 (start) 0 27533 consensus 0.2468674 NA
#> 2 consensus 1 6329 plan 0.3957971 0.3340302623
#> 3 plan 1 6157 plan 0.3742082 0.2341431253
#> 4 discuss 1 3951 consensus 0.3211845 0.5556255176
#> 5 emotion 1 2837 cohesion 0.3253437 0.5551452430
#> 6 consensus -> plan 2 2336 plan 0.3754281 0.0006484903
#> 7 plan -> plan 2 2108 plan 0.3757116 0.0004321472
#> 8 coregulate 1 1970 discuss 0.2736041 0.1808905912
#> changes_prediction
#> 1 NA
#> 2 TRUE
#> 3 TRUE
#> 4 FALSE
#> 5 TRUE
#> 6 FALSE
#> 7 FALSE
#> 8 TRUE
divergent_pathways(tree, top = 8) # where adding history changes the prediction most
#> pathway depth count likely_next next_probability
#> 1 synthesis -> discuss -> consensus 3 10 coregulate 0.5956000
#> 2 consensus -> cohesion -> plan 3 12 plan 0.8268333
#> 3 synthesis 1 652 consensus 0.4630613
#> 4 cohesion -> discuss -> emotion 3 10 cohesion 0.4965000
#> 5 adapt 1 509 consensus 0.4741100
#> 6 monitor -> monitor -> discuss 3 12 discuss 0.2487500
#> 7 cohesion -> cohesion -> consensus 3 19 plan 0.3661053
#> 8 coregulate -> emotion -> monitor 3 13 consensus 0.3821538
#> divergence changes_prediction
#> 1 0.9397936 TRUE
#> 2 0.9259734 FALSE
#> 3 0.8917924 FALSE
#> 4 0.8716723 TRUE
#> 5 0.7892874 FALSE
#> 6 0.7829439 TRUE
#> 7 0.7560378 FALSE
#> 8 0.7523873 TRUE
sharp_pathways(tree, top = 8) # the most peaked next-action predictions
#> pathway depth count likely_next next_probability
#> 1 consensus -> cohesion -> plan 3 12 plan 0.8268333
#> 2 discuss -> coregulate -> cohesion 3 11 consensus 0.7217273
#> 3 coregulate -> coregulate -> plan 3 14 plan 0.7088571
#> 4 consensus -> adapt -> consensus 3 12 plan 0.6616667
#> 5 cohesion -> discuss -> synthesis 3 12 consensus 0.6616667
#> 6 emotion -> discuss -> cohesion 3 14 consensus 0.6380714
#> 7 discuss -> plan -> plan 3 11 consensus 0.6316364
#> 8 emotion -> emotion -> consensus 3 58 plan 0.6161034
#> divergence changes_prediction
#> 1 0.9259734 FALSE
#> 2 0.5087962 FALSE
#> 3 0.6616111 FALSE
#> 4 0.3790714 FALSE
#> 5 0.3634509 FALSE
#> 6 0.2248026 FALSE
#> 7 0.5802868 TRUE
#> 8 0.2405929 FALSE
Read the divergent table in two layers. The very top rows can have
large divergence on a small count – a short
history seen just over the min_count floor that happened to
resolve one way. Those are small-sample mirages; the bootstrap in
section 7 exists to disarm them. The rows that also carry a
large count are the well-supported redirections worth
quoting.
The sharp table teaches the same caution from the probability side: a
next_probability near 1 on a low count is a
near-empty cell after smoothing, not a law of behaviour. Sharpness
with support is a rule; sharpness without it is
noise.
5. Per-context diagnostics
tree_dependence() is the information-theoretic
decomposition the KL pruning rule thresholds: per context, how many
bits of next-action uncertainty the extra history
removes (entropy_drop), and whether it flips the modal
prediction.
tree_dependence(tree, sort_by = "entropy_drop", top = 8)
#> pathway depth count divergence entropy
#> 1 consensus -> cohesion -> plan 3 12 0.9259734 0.885185
#> 2 cohesion -> discuss -> discuss 3 13 0.6501997 1.599492
#> 3 synthesis -> discuss -> consensus 3 10 0.9397936 1.358019
#> 4 coregulate -> synthesis -> consensus 3 14 0.3945507 1.440219
#> 5 coregulate -> coregulate -> plan 3 14 0.6616111 1.347984
#> 6 discuss -> coregulate -> cohesion 3 11 0.5087962 1.160730
#> 7 monitor -> discuss -> monitor 3 13 0.4218380 1.520958
#> 8 monitor -> cohesion -> plan 3 10 0.4461585 1.432416
#> entropy_before entropy_drop likely_next likely_before changes_prediction
#> 1 2.289429 1.4042437 plan plan FALSE
#> 2 2.683111 1.0836194 consensus consensus FALSE
#> 3 2.344887 0.9868676 coregulate plan TRUE
#> 4 2.383187 0.9429680 plan plan FALSE
#> 5 2.276847 0.9288633 plan plan FALSE
#> 6 2.067552 0.9068226 consensus consensus FALSE
#> 7 2.401484 0.8805256 discuss discuss FALSE
#> 8 2.289429 0.8570129 consensus plan TRUE
A large entropy_drop with
changes_prediction = TRUE is the most valuable kind of
context: it both sharpens and redirects. Watch for
negative entropy_drop – the longer history
left the next action more uncertain than its parent; that is
the textbook signature of a context pruning should remove.
6. Prune to the reliable tree
pruned <- prune_tree(tree, criterion = "G2", alpha = 0.05)
pruned
#> <transitiontrees> 23 nodes, depth <= 3, 9 states [pruned]
#> alphabet : adapt, cohesion, consensus, coregulate, discuss, emotion, monitor, plan, synthesis
#> fit on : 2000 sequences, 27533 observations
#> smoothing: floor(ymin=0.001, rule=interpolate) min_count = 10
#> pruned by: G2 alpha = 0.05
#> (start) n=27533 -> consensus (0.25)
#> |-- adapt n=509 -> consensus (0.47)
#> |-- cohesion n=1695 -> consensus (0.50)
#> | `-- cohesion n=45 -> consensus (0.42)
#> |-- consensus n=6329 -> plan (0.40)
#> | |-- cohesion n=795 -> plan (0.38)
#> | | `-- cohesion n=19 -> plan (0.37)
#> | `-- emotion n=830 -> plan (0.39)
#> | `-- emotion n=58 -> plan (0.62)
#> |-- coregulate n=1970 -> discuss (0.27)
#> |-- discuss n=3951 -> consensus (0.32)
#> | |-- adapt n=29 -> adapt (0.24)
#> | `-- coregulate n=486 -> consensus (0.32)
#> | `-- discuss n=88 -> consensus (0.27)
#> |-- emotion n=2837 -> cohesion (0.33)
#> | |-- emotion n=199 -> cohesion (0.35)
#> | `-- plan n=831 -> consensus (0.33)
#> | `-- cohesion n=33 -> cohesion (0.27)
#> |-- monitor n=1433 -> discuss (0.38)
#> |-- plan n=6157 -> plan (0.37)
#> | `-- cohesion n=221 -> plan (0.36)
#> | `-- consensus n=12 -> plan (0.83)
#> `-- synthesis n=652 -> consensus (0.46)
The pruned banner reports the surviving node count and criterion;
compare it to the unpruned tree from section 2. Each
removed context failed a likelihood-ratio G-squared test against its
one-shorter parent: the extra history did not explain enough added
variation in the next action to justify keeping it. That the tree
collapses so far is itself a finding – most of the grown depth was
unsupported, and the durable structure lives near the root.
7. Held-out predictive quality
The honest, out-of-sample estimate comes from cross-validation, which
tune_tree() runs at the sequence level over a
(max_depth, min_count, ...) grid – no hand-made train/test
split. The in-sample perplexity is the optimistic bound; the
cross-validated winner is the figure to report.
model_fit(pruned)$perplexity # in-sample (optimistic)
#> [1] 5.427279
tg <- tune_tree(group_regulation_long,
actor = "Actor", time = "Time", action = "Action",
max_depth = 1L:3L, min_count = 10L, folds = 5L, seed = 1L)
attr(tg, "best") # cross-validated winner
#> max_depth nmin smoothing prune logLik n_scored
#> 1 1 10 floor(ymin=0.001, rule=interpolate) FALSE -46738.34 27533
#> perplexity n_nodes_avg folds_failed
#> 1 5.460492 10 0
A cross-validated perplexity close to the in-sample value is the
signature of a well-pruned model that generalises; a large gap would say
prune harder.
mine_sequences() then surfaces the sessions the fitted
model predicts worst – the atypical regulation trajectories worth a
closer look – and score_positions() the individual moves it
is most blindsided by:
wide <- prepare_input(group_regulation_long,
actor = "Actor", time = "Time", action = "Action")
mine_sequences(pruned, newdata = wide, which = "surprising", n = 5L)
#> sequence_id n_scored log_lik perplexity
#> 1 1559 2 -8.349842 65.03470
#> 2 446 2 -6.908739 31.63833
#> 3 1823 3 -9.619186 24.68993
#> 4 1323 3 -9.542743 24.06875
#> 5 1671 3 -9.140954 21.05177
score_positions(pruned, newdata = wide, worst = 5L)
#> sequence_id position matched_context observed predicted_prob log_lik
#> 1 69 22 plan adapt 0.0009745006 -6.933585
#> 2 235 17 plan adapt 0.0009745006 -6.933585
#> 3 974 20 plan adapt 0.0009745006 -6.933585
#> 4 1227 7 plan adapt 0.0009745006 -6.933585
#> 5 1424 3 plan adapt 0.0009745006 -6.933585
8. Bootstrap reliability
prune_tree() asked “which contexts pass a criterion
in this dataset?”. The bootstrap asks the stricter question –
“which pass reproducibly under resampling?” – and reports two
flags. stable: the count reproduces.
informative: the G-squared against the
parent reproducibly clears the chi-square bar. A claim worth making is
both.
boot <- bootstrap_pathways(pruned, iter = 100L, stat = "count", seed = 1L)
boot
#> <transitiontrees_bootstrap> 100 resamples
#> stability : count in [0.50, 1.50] x observed, p < 0.05
#> informative: G^2 > qchisq(0.95, df=k-1) = 15.51, threshold 0.80
#> pathways : 23 total, 21 stable, 14 informative, 13 both
#>
#> top pathways (stable + informative first):
#> pathway depth count p_stability stability_rate stable
#> consensus 1 6329 0.01 1 TRUE
#> plan 1 6157 0.01 1 TRUE
#> discuss 1 3951 0.01 1 TRUE
#> emotion 1 2837 0.01 1 TRUE
#> coregulate 1 1970 0.01 1 TRUE
#> cohesion 1 1695 0.01 1 TRUE
#> monitor 1 1433 0.01 1 TRUE
#> synthesis 1 652 0.01 1 TRUE
#> adapt 1 509 0.01 1 TRUE
#> discuss -> coregulate -> discuss 3 88 0.01 1 TRUE
#> informative_rate informative mean_G2 ci_G2_lo ci_G2_hi
#> 1.00 TRUE 2953.021 2734.523 3152.769
#> 1.00 TRUE 1999.111 1875.164 2148.083
#> 1.00 TRUE 3052.763 2925.134 3219.423
#> 1.00 TRUE 2201.118 2025.104 2349.157
#> 1.00 TRUE 502.777 428.421 600.107
#> 1.00 TRUE 750.700 662.513 854.702
#> 1.00 TRUE 460.401 391.198 557.597
#> 1.00 TRUE 840.484 724.436 972.445
#> 1.00 TRUE 580.569 508.107 674.837
#> 0.96 TRUE 25.679 15.060 42.570
#> # ... 13 more pathways (use summary(x) for full table)
head(summary(boot), 10)
#> pathway depth count likely_next next_probability
#> 1 consensus 1 6329 plan 0.3957971
#> 2 plan 1 6157 plan 0.3742082
#> 3 discuss 1 3951 consensus 0.3211845
#> 4 emotion 1 2837 cohesion 0.3253437
#> 5 coregulate 1 1970 discuss 0.2736041
#> 6 cohesion 1 1695 consensus 0.4979351
#> 7 monitor 1 1433 discuss 0.3754361
#> 8 synthesis 1 652 consensus 0.4662577
#> 9 adapt 1 509 consensus 0.4774067
#> 10 discuss -> coregulate -> discuss 3 88 consensus 0.2727273
#> divergence changes_prediction G2 p_stability stability_rate stable
#> 1 0.3340303 TRUE 2930.7338 0.00990099 1 TRUE
#> 2 0.2341431 TRUE 1998.5086 0.00990099 1 TRUE
#> 3 0.5556255 FALSE 3043.2993 0.00990099 1 TRUE
#> 4 0.5551452 TRUE 2183.3402 0.00990099 1 TRUE
#> 5 0.1808906 TRUE 494.0122 0.00990099 1 TRUE
#> 6 0.3149856 FALSE 740.1435 0.00990099 1 TRUE
#> 7 0.2283039 TRUE 453.5394 0.00990099 1 TRUE
#> 8 0.9091915 FALSE 821.7854 0.00990099 1 TRUE
#> 9 0.8132687 FALSE 573.8618 0.00990099 1 TRUE
#> 10 0.1663148 FALSE 20.2894 0.00990099 1 TRUE
#> informative_rate informative flip_consistency mean_count sd_count
#> 1 1.00 TRUE 0.92 6342.99 109.57655
#> 2 1.00 TRUE 0.92 6161.66 130.96821
#> 3 1.00 TRUE 0.92 3958.31 81.97335
#> 4 1.00 TRUE 0.67 2836.33 61.00390
#> 5 1.00 TRUE 0.99 1972.85 53.81062
#> 6 1.00 TRUE 0.92 1696.68 42.70970
#> 7 1.00 TRUE 1.00 1430.31 40.39144
#> 8 1.00 TRUE 0.92 658.20 26.31357
#> 9 1.00 TRUE 0.92 511.36 20.96390
#> 10 0.96 TRUE 0.80 87.01 10.51838
#> ci_count_lo ci_count_hi mean_next_probability sd_next_probability
#> 1 6115.925 6536.925 0.3959500 0.006381042
#> 2 5857.650 6366.175 0.3732453 0.006934701
#> 3 3838.900 4157.050 0.3202636 0.006807873
#> 4 2735.600 2952.575 0.3301679 0.006276168
#> 5 1875.275 2073.575 0.2737128 0.010252119
#> 6 1611.425 1767.675 0.4979885 0.011727875
#> 7 1356.850 1511.625 0.3754592 0.013158816
#> 8 609.850 716.000 0.4691459 0.018529368
#> 9 466.425 544.150 0.4788970 0.021643330
#> 10 68.475 109.050 0.2821353 0.041049294
#> ci_next_probability_lo ci_next_probability_hi mean_divergence sd_divergence
#> 1 0.3810029 0.4070170 0.3358281 0.010620980
#> 2 0.3598353 0.3869965 0.2340653 0.008670898
#> 3 0.3061619 0.3330049 0.5564292 0.014373664
#> 4 0.3198049 0.3451925 0.5598475 0.021406786
#> 5 0.2546615 0.2903713 0.1837969 0.015985660
#> 6 0.4750867 0.5160186 0.3191616 0.020813645
#> 7 0.3547976 0.4010721 0.2322337 0.021001349
#> 8 0.4384477 0.5138353 0.9209807 0.058113430
#> 9 0.4372385 0.5240070 0.8192246 0.051354931
#> 10 0.2117708 0.3637233 0.2138777 0.056185496
#> ci_divergence_lo ci_divergence_hi mean_G2 sd_G2 ci_G2_lo
#> 1 0.3148910 0.3529989 2953.02132 106.335090 2734.52287
#> 2 0.2178176 0.2533216 1999.11080 79.507861 1875.16412
#> 3 0.5348465 0.5860827 3052.76257 81.500119 2925.13359
#> 4 0.5151091 0.5996423 2201.11768 91.903624 2025.10387
#> 5 0.1588430 0.2133030 502.77714 47.114738 428.42086
#> 6 0.2866160 0.3609771 750.69969 52.467915 662.51281
#> 7 0.1978436 0.2723805 460.40086 43.010781 391.19790
#> 8 0.8041894 1.0322695 840.48398 64.700502 724.43592
#> 9 0.7347244 0.9244516 580.56882 41.164021 508.10688
#> 10 0.1270018 0.3397060 25.67886 7.081377 15.06023
#> ci_G2_hi
#> 1 3152.76930
#> 2 2148.08313
#> 3 3219.42286
#> 4 2349.15734
#> 5 600.10734
#> 6 854.70155
#> 7 557.59747
#> 8 972.44484
#> 9 674.83712
#> 10 42.57021
summary() sorts the trustworthy (stable and
informative) pathways first, so the top rows are the defensible set. The
two flags screen different failure modes. stable alone
keeps high-count noise pathways; informative alone could
surface a low-count borderline pathway whose sample G-squared is high by
chance. Their conjunction is the defensible set.
plot(boot)
#> `height` was translated to `width`.
In the forest plot each bar is a 95% bootstrap interval on G-squared;
the dashed line is the chi-square critical value. A bar entirely to the
right is reproducibly informative; a bar straddling the line is not safe
to claim.
9. Do high and low achievers regulate differently?
Fit one tree per group in a single call with
group =, then test where the groups diverge with a
permutation null. The grouping variable is an external student
attribute (Achiever), not derived from the actions
themselves – otherwise the comparison would be circular.
grp <- context_tree(group_regulation_long,
actor = "Actor", time = "Time", action = "Action",
group = "Achiever", max_depth = 2L, min_count = 10L)
cmp <- compare_groups(grp, iter = 199L, seed = 1L)
cmp$omnibus
#> axis statistic value p_value
#> 1 behavioral count-weighted JSD (bits) 1772.142 0.005
#> 2 usage sum G^2 1356.465 0.005
The omnibus table reports two axes. behavioral is
the count-weighted Jensen-Shannon divergence (bits) between the groups’
next-action distributions, summed over shared contexts – “given the same
history, do the groups do different things next?”.
usage is the summed G-squared homogeneity statistic –
“do they reach the same contexts at the same rates?”. Each
p_value comes from permuting the group labels.
plot_difference(grp, depth = 1L)
The per-context residual map shows where the groups differ:
red and blue cells are the contexts a high achiever and a low achiever
resolve toward different next actions. depth = 1L restricts
it to the single-action contexts so the rows stay readable; drop it (or
raise it) to inspect deeper histories.
Synthesis
Pulling the thread through every section:
- The action alphabet is imbalanced – frequency is a
misleading lens and modal predictions are trivially
consensus/plan.
- Memory is short – pruning collapses the tree to a
small set of contexts, and held-out perplexity confirms the shallow
model generalises.
- The insight is in the divergent, well-counted
contexts – not the common ones, and not the spectacular
low-count tail.
- Only the stable-and-informative pathways are
claimable – the bootstrap is the trust filter between an
eyeballed table and a finding.
- High and low achievers regulate measurably
differently – the permutation test licenses the claim that the
omnibus statistic is real, not a relabelling artefact.
Each claim is anchored to a function whose output you can re-run –
the whole point of a pathway-centric, testable model.