1. Data

Table ST1. Repositories per domain, before and after filtering

Domain Raw After filtering
open_source 162730 10864
bioinformatics 39539 1493
astrophysics 3254 243
software_engineering 95429 1088
image_recognition 39912 149

2. RQI Scoring

RQI sub-scores use piecewise-linear interpolation between fixed anchor points, so scores do not depend on the sample.

\[ \text{RQI}_i = \frac{1}{3}\Big( f_{\text{recency}}(\text{days\_since\_last\_commit}_i) + f_{\text{issues}}(\text{median\_issue\_days\_open}_i) + f_{\text{popularity}}\big(\log_{10}(1 + \text{stars}_i + \text{forks}_i + \text{watchers}_i)\big) \Big) \]

where each \(f\) interpolates between fixed anchors on a 0–5 scale. Values beyond the first or last anchor take the score of that anchor.

Table ST2. RQI scoring anchors. Scores are linearly interpolated between anchors.

\[ \begin{array}{l|cccccccc} \text{Recency: days since last commit} & 0 & 14 & 30 & 60 & 90 & 180 & 365 & 730 \\ \hline f_{\text{recency}} & 5.0 & 4.5 & 4.0 & 3.5 & 3.0 & 2.0 & 1.3 & 0.0 \end{array} \]

\[ \begin{array}{l|ccccccc} \text{Backlog: median days issues stay open} & 0 & 7 & 14 & 30 & 60 & 120 & 365 \\ \hline f_{\text{issues}} & 5.0 & 4.5 & 4.0 & 3.5 & 3.0 & 2.0 & 1.0 \end{array} \]

\[ \begin{array}{l|ccccc} \text{Popularity: } \log_{10}(1 + \text{stars} + \text{forks} + \text{watchers}) & 0 & 1 & 2 & 3 & 4 \\ \hline f_{\text{popularity}} & 1.0 & 2.0 & 3.0 & 4.0 & 5.0 \end{array} \]

Missing values: a repository with no issue-closure median is scored as 365 days open (backlog 1.0).

3. Summary Statistics

3.1 Overall and by Type

Table ST3. Overall repository summary

Statistic Value
Repositories 13,670
Without CI 5,085
Without tests 1,742
CI coverage (%) 62.8
Test coverage (%) 87.3
Neither CI nor tests (%) 7.5
Mean RQI 2.88
SD of RQI 0.90
Minimum RQI 0.67
Maximum RQI 5.00

Table ST4. Summary by repository type

Type n CI (%) Tests (%) Neither (%) Mean RQI SD RQI
astrophysics 238 73.9 92.4 4.6 2.97 0.84
open_source 10773 64.4 86.8 7.3 2.96 0.92
bioinformatics 1456 65.9 89.3 7.7 2.79 0.78
image_recognition 144 36.1 77.8 18.1 2.56 0.74
software_engineering 1059 43.6 89.0 8.6 2.22 0.61

4. Analysis of Normality for RQI

4.1 RQI by Repository Type

Fig S1. RQI distribution by repository type. Stacked histogram with density overlays.

4.2 Overall RQI Distribution

Fig S2. Overall RQI distribution and Q–Q plot. Dashed line marks the mean.

Table ST5. Overall RQI distribution: summary statistics and normality diagnostics

n Min Q1 Median Mean Q3 Max SD Skewness KS D KS p
13670 0.667 2.217 2.715 2.878 3.55 5 0.904 0.338 0.073 <0.001

4.3 RQI Distribution by Type

Fig S3. RQI distribution by repository type. Kolmogorov–Smirnov statistics and skewness shown per type.

Table ST6. Normality diagnostics for RQI by repository type

Type n KS D KS p Skewness
astrophysics 238 0.082 0.084 -0.191
bioinformatics 1456 0.055 <0.001 0.170
image_recognition 144 0.064 0.603 0.311
open_source 10773 0.071 <0.001 0.278
software_engineering 1059 0.141 <0.001 0.677

P values indicate non-normality, but the plots and low D values suggest normality is adequate for ANOVA.

5. ANOVA and Tukey Tests

Table ST7. One-way ANOVA of RQI by repository type

Term Df Sum Sq Mean Sq F p
Repository type 4 555.98 138.99 178.97 <0.001
Residuals 13665 10612.86 0.78

Table ST8. Tukey HSD pairwise comparisons

Comparison Difference CI lower CI upper Adj. p
bioinformatics vs. astrophysics -0.185 -0.353 -0.017 0.022
image_recognition vs. astrophysics -0.409 -0.663 -0.155 <0.001
open_source vs. astrophysics -0.017 -0.175 0.140 0.998
software_engineering vs. astrophysics -0.756 -0.928 -0.583 <0.001
image_recognition vs. bioinformatics -0.224 -0.434 -0.014 0.030
open_source vs. bioinformatics 0.168 0.101 0.235 <0.001
software_engineering vs. bioinformatics -0.571 -0.668 -0.474 <0.001
open_source vs. image_recognition 0.392 0.190 0.594 <0.001
software_engineering vs. image_recognition -0.347 -0.560 -0.133 <0.001
software_engineering vs. open_source -0.739 -0.816 -0.661 <0.001

6. Significant Comparisons

Table ST9. Tukey HSD comparisons with adjusted p < 0.05

Group1 Group2 Difference CI lower CI upper Adj. p
bioinformatics astrophysics -0.185 -0.353 -0.017 0.022
image_recognition astrophysics -0.409 -0.663 -0.155 <0.001
software_engineering astrophysics -0.756 -0.928 -0.583 <0.001
image_recognition bioinformatics -0.224 -0.434 -0.014 0.030
open_source bioinformatics 0.168 0.101 0.235 <0.001
software_engineering bioinformatics -0.571 -0.668 -0.474 <0.001
open_source image_recognition 0.392 0.190 0.594 <0.001
software_engineering image_recognition -0.347 -0.560 -0.133 <0.001
software_engineering open_source -0.739 -0.816 -0.661 <0.001

7. Mean Scores by Type

Table 1 (main text). Mean scores by repository type.

8. Group Means

Fig S4. Mean RQI, bioinformatics vs other domains. Means with 95% CIs; brackets give Tukey-adjusted p values for significant comparisons.

9. Radar Plots

9.1 Top, Median, Bottom 100

Fig 1 (main text). Radar plots of the top, median, and bottom 100 repositories by RQI. Panels a–c; values scaled 0–100 (100 = best).

Table ST10. Fig 1 values (0–100).

Radar plot values for Fig 1 (0-100 scale; outer edge = 100 = best).
Group Recent Activity Backlog Health Popularity
Top 100 (highest RQI) 100.0 86.5 77.2
Median 100 81.8 33.8 36.0
Bottom 100 (lowest RQI) 69.0 0.2 7.9

9.2 Bioinformatics Top, Median, Bottom 100

Fig S5. Radar plots of the top, median, and bottom 100 bioinformatics repositories.

Table ST11. Fig S5 values (0–100).

Radar plot values for Fig S5 (bioinformatics, 0-100 scale).
Group Recent Activity Backlog Health Popularity
Top 100 (highest RQI) 99.8 39.8 43.4
Median 100 89.0 25.7 30.7
Bottom 100 (lowest RQI) 74.9 0.8 21.1

9.3 Profile by Domain

Fig S6. Average RQI component profile by domain.

Table ST12. Fig S6 values (0–100).

Radar plot values for Fig S6, by domain (0-100 scale).
Domain Recent Activity Backlog Health Popularity
astrophysics 92.1 20.9 25.9
bioinformatics 88.1 20.7 28.6
image_recognition 78.9 31.6 30.5
open_source 86.3 29.3 35.2
software_engineering 78.3 33.3 11.9

9.4 Top vs Bottom 20 by Domain

Fig S7. Top 20 vs bottom 20 repositories by domain. Left: highest RQI; right: lowest.

Table ST13. Fig S7 values (0–100).

Radar plot values for Fig S7, top 20 vs bottom 20 by domain (0-100 scale).
Domain Group Recent Activity Backlog Health Popularity
astrophysics Top 20 99.9 39.9 39.2
astrophysics Bottom 20 81.5 1.1 20.3
bioinformatics Top 20 100.0 45.6 52.6
bioinformatics Bottom 20 68.4 0.3 14.4
image_recognition Top 20 98.2 47.7 42.5
image_recognition Bottom 20 74.3 2.6 18.5
open_source Top 20 100.0 100.0 80.2
open_source Bottom 20 64.4 0.2 4.0
software_engineering Top 20 99.7 46.3 35.5
software_engineering Bottom 20 68.4 0.1 5.2

10. CI and Test Prevalence by Quality

10.1 CI and Tests by RQI

Fig 2 (main text). CI and testing prevalence by RQI. Bins of 0.15 RQI (n ≥ 5); point size shows bin n; lines are OLS fits.

10.2 Repos per RQI Bin

Table ST14. Repositories per RQI bin (bin width 0.15)

RQI bin Repositories
0.60 11
0.75 27
0.90 100
1.05 92
1.20 128
1.35 136
1.50 184
1.65 265
1.80 476
1.95 668
2.10 997
2.25 1161
2.40 1036
2.55 944
2.70 979
2.85 782
3.00 675
3.15 552
3.30 526
3.45 444
3.60 520
3.75 466
3.90 454
4.05 408
4.20 403
4.35 366
4.50 301
4.65 232
4.80 159
4.95 178

10.3 CI and Tests by Quality Tier

Fig S8. CI and testing prevalence by RQI tier. Low (RQI ≤ 2), mid, and high (RQI ≥ 4) tiers; Holm-adjusted pairwise tests shown.

Table ST15. Holm-adjusted pairwise tests of prevalence between RQI tiers (p values)

Comparison CI Testing
2 vs. 1 <0.001 0.011
3 vs. 1 <0.001 <0.001
3 vs. 2 <0.001 <0.001

11. CI and Tests by Type

11.1 t-Tests

Table ST16. Welch two-sample t-tests of RQI by CI and test presence

Comparison Mean 1 Mean 2 Difference t df p CI lower CI upper
RQI by CI (1 = no CI, 2 = has CI) 2.518 3.090 -0.572 -40.01 12676.2 <0.001 -0.601 -0.544
RQI by tests (1 = no tests, 2 = has tests) 2.628 2.914 -0.286 -13.95 2498.1 <0.001 -0.326 -0.245

11.2 Mean RQI by CI and Tests

Fig S9. Mean RQI by CI and testing presence. Means ± SE; brackets give Welch t-test p values.

11.3 Within-Type Tests

Table ST17. Mean RQI with and without CI or tests, by repository type (Welch t-test p values)

Type With CI No CI p (CI) With tests No tests p (tests)
astrophysics 3.16 2.45 <0.001 3.02 2.45 0.003
bioinformatics 2.92 2.53 <0.001 2.82 2.52 <0.001
image_recognition 2.80 2.43 0.008 2.62 2.37 0.100
open_source 3.17 2.58 <0.001 3.00 2.68 <0.001
software_engineering 2.29 2.16 <0.001 2.22 2.22 0.930

11.4 CI Effect by Type

Fig S10. Mean RQI by CI presence, within each type. Means ± SE; brackets give Welch t-test p values.

11.5 Testing Effect by Type

Fig S11. Mean RQI by testing presence, within each type. Means ± SE; brackets give Welch t-test p values.

12. Languages

12.1 Language List

Table ST18. Most frequent languages.

Language Repositories
JavaScript 6059
Shell 5501
HTML 5499
Python 5152
CSS 5104
Dockerfile 3009
TypeScript 2436
Makefile 2057
C++ 1879
C 1807
Java 1661
SCSS 1106
Batchfile 938
CMake 920
C# 866
PHP 794
Ruby 762
Jupyter Notebook 718
Go 692
Vue 444
Kotlin 443
R 441
Rust 410
Objective-C 405
TeX 397
PowerShell 363
Perl 359
Swift 359
PLpgSQL 286
Smarty 252

12.2 Primary Language

Primary language is the one with the most bytes.

Table ST19. Most frequent primary languages

Language Repositories
Python 2572
JavaScript 1635
TypeScript 1505
Java 1050
C++ 945
HTML 624
C# 611
C 449
PHP 443
Go 413
Jupyter Notebook 354
Rust 226
CSS 212
Kotlin 175
Shell 174
Ruby 166
Dart 161
R 159
Vue 144
Swift 94
TeX 72
Julia 50
Nextflow 49
Scala 49
Svelte 43
SCSS 42
Objective-C 39
Perl 35
Elixir 34
GDScript 33

12.3 RQI by Language

Table ST20. Mean RQI by primary language (languages with n ≥ 20; top 30 by mean RQI)

Language n Mean RQI SD RQI CI (%) Tests (%)
MDX 28 3.52 0.79 85.7 96.4
Rust 226 3.30 0.95 88.5 90.7
Go 413 3.26 0.96 86.9 91.8
Fortran 27 3.22 0.84 77.8 96.3
Pascal 25 3.19 0.99 56.0 72.0
TypeScript 1505 3.15 0.96 79.8 94.0
GDScript 33 3.12 1.01 54.5 90.9
C 449 3.11 0.91 54.8 86.0
Julia 50 3.06 0.84 92.0 98.0
Dart 161 3.01 0.88 70.2 99.4
Kotlin 175 3.00 0.96 76.6 100.0
Shell 174 3.00 0.89 66.1 73.0
C++ 945 2.99 0.91 60.2 89.8
Vue 144 2.97 0.95 66.7 89.6
MATLAB 28 2.96 0.95 35.7 89.3
Python 2572 2.95 0.85 66.6 88.5
Swift 94 2.92 0.84 61.7 98.9
Svelte 43 2.90 0.97 67.4 90.7
Elixir 34 2.90 1.01 85.3 97.1
Lua 31 2.87 0.97 51.6 83.9
C# 611 2.86 0.94 56.3 89.4
Nextflow 49 2.85 0.76 81.6 95.9
PHP 443 2.85 0.94 57.1 91.6
Dockerfile 22 2.82 1.10 81.8 54.5
Jupyter Notebook 354 2.78 0.80 52.3 87.6
R 159 2.75 0.77 63.5 89.3
Perl 35 2.75 0.77 42.9 91.4
Scala 49 2.74 0.90 89.8 100.0
HTML 624 2.66 0.84 47.0 76.8
Java 1050 2.66 0.88 55.0 96.5

Note: JavaScript is not shown because it falls outside the top 30 by mean RQI (rank 33 of 37 languages with n ≥ 20). Its values: n = 1635, mean RQI = 2.61, SD = 0.81, CI = 56.9%, tests = 84.5%.

12.4 RQI by Language, CI and Tests

Table ST21. Mean RQI by primary language, CI and tests (languages with n ≥ 20; top 30 by mean RQI). Mean RQI (0–5) by presence of tests and CI. JavaScript is not shown because it falls outside the top 30 by mean RQI (rank 33 of 37). Its values: n = 1635, mean RQI = 2.61; no tests = 2.63, has tests = 2.61; no CI = 2.42, has CI = 2.75.

Language n Mean RQI No tests Has tests No CI Has CI
MDX 28 3.52 4.19 3.49 3.37 3.54
Rust 226 3.30 2.92 3.34 2.78 3.37
Go 413 3.26 3.10 3.27 2.65 3.35
Fortran 27 3.22 4.17 3.18 2.86 3.32
Pascal 25 3.19 2.92 3.30 2.91 3.42
TypeScript 1505 3.15 2.65 3.19 2.74 3.26
GDScript 33 3.12 1.96 3.24 2.75 3.43
C 449 3.11 2.67 3.18 2.76 3.39
Julia 50 3.06 2.96 3.06 2.80 3.08
Dart 161 3.01 4.47 3.00 2.65 3.17
Kotlin 175 3.00 3.00 2.57 3.13
Shell 174 3.00 2.87 3.04 2.64 3.18
C++ 945 2.99 2.62 3.03 2.58 3.26
Vue 144 2.97 2.41 3.04 2.61 3.15
MATLAB 28 2.96 2.35 3.03 2.58 3.63
Python 2572 2.95 2.71 2.98 2.66 3.10
Swift 94 2.92 1.98 2.93 2.44 3.22
Svelte 43 2.90 2.24 2.97 2.36 3.16
Elixir 34 2.90 1.68 2.93 2.50 2.96
Lua 31 2.87 2.74 2.90 2.54 3.18
C# 611 2.86 2.48 2.91 2.45 3.18
Nextflow 49 2.85 2.05 2.88 2.38 2.95
PHP 443 2.85 2.40 2.89 2.46 3.14
Dockerfile 22 2.82 2.83 2.82 2.16 2.97
Jupyter Notebook 354 2.78 2.63 2.80 2.52 3.02
R 159 2.75 2.23 2.81 2.47 2.91
Perl 35 2.75 2.35 2.78 2.61 2.93
Scala 49 2.74 2.74 2.64 2.75
HTML 624 2.66 2.52 2.70 2.40 2.94
Java 1050 2.66 2.50 2.66 2.34 2.92

Fig S12. Mean RQI by language and CI presence. Languages with n ≥ 20; means ± SE.

Fig S13. Mean RQI by language and testing presence. Languages with n ≥ 20; means ± SE.

Fig S14. Mean RQI by language and CI/testing group. Languages with n ≥ 20.

12.5 Bioinformatics Languages

Table ST22. Most frequent primary languages, bioinformatics

Language Repositories
Python 585
R 118
C++ 88
Jupyter Notebook 80
HTML 72
JavaScript 63
Nextflow 49
Java 48
Rust 46
C 42
Shell 27
Julia 26
Perl 23
TypeScript 19
TeX 15
Scala 14
Go 12
Ruby 10
Cython 6
D 6
WDL 6
Groovy 5
MATLAB 5
Vue 4
C# 3
CSS 3
Clojure 3
Kotlin 3
OCaml 3
PHP 3

Table ST23. Most frequent languages, bioinformatics

Language Repositories
Python 978
Shell 712
HTML 376
Dockerfile 350
R 321
Makefile 301
JavaScript 252
CSS 235
C++ 177
Jupyter Notebook 176
C 159
Perl 123
TeX 114
Nextflow 88
Java 77
Rust 59
CMake 56
Batchfile 49
SCSS 45
TypeScript 42
Cython 34
Roff 31
Ruby 31
Julia 30
Groovy 25
WDL 23
Scala 22
Go 19
M4 18
Awk 16

Table ST24. Mean RQI by primary language, bioinformatics (languages with n ≥ 10; top 30 by mean RQI)

Language n Mean RQI SD RQI CI (%) Tests (%)
Go 12 3.15 0.79 91.7 100.0
Rust 46 3.06 0.68 84.8 87.0
C 42 3.04 0.74 50.0 81.0
TypeScript 19 2.96 0.80 89.5 94.7
Java 48 2.88 0.92 62.5 97.9
Julia 26 2.87 0.86 84.6 96.2
Python 585 2.85 0.76 73.3 92.1
Nextflow 49 2.85 0.76 81.6 95.9
C++ 88 2.80 0.66 63.6 92.0
JavaScript 63 2.77 0.71 65.1 92.1
R 118 2.76 0.75 62.7 88.1
Shell 27 2.74 0.88 37.0 74.1
Ruby 10 2.65 0.99 80.0 100.0
Jupyter Notebook 80 2.59 0.71 42.5 83.8
Perl 23 2.58 0.65 34.8 91.3
Scala 14 2.57 1.01 100.0 100.0
HTML 72 2.56 0.86 51.4 83.3
TeX 15 2.39 0.66 53.3 66.7

Table ST25. Mean RQI by primary language, CI and tests, bioinformatics (languages with n ≥ 10; top 30 by mean RQI). Mean RQI (0–5) by presence of tests and CI.

Language n Mean RQI No tests Has tests No CI Has CI
Go 12 3.15 3.15 3.76 3.10
Rust 46 3.06 2.85 3.09 2.94 3.08
C 42 3.04 2.85 3.08 2.86 3.22
TypeScript 19 2.96 2.27 2.99 3.29 2.92
Java 48 2.88 2.23 2.90 2.64 3.03
Julia 26 2.87 2.96 2.86 2.80 2.88
Python 585 2.85 2.62 2.87 2.58 2.95
Nextflow 49 2.85 2.05 2.88 2.38 2.95
C++ 88 2.80 2.92 2.79 2.55 2.94
JavaScript 63 2.77 2.46 2.79 2.67 2.82
R 118 2.76 2.23 2.83 2.55 2.88
Shell 27 2.74 2.80 2.71 2.59 2.99
Ruby 10 2.65 2.65 2.32 2.74
Jupyter Notebook 80 2.59 2.35 2.64 2.51 2.70
Perl 23 2.58 2.14 2.62 2.55 2.63
Scala 14 2.57 2.57 2.57
HTML 72 2.56 2.37 2.60 2.23 2.87
TeX 15 2.39 2.34 2.42 2.47 2.32

Fig S15. Mean RQI by language, bioinformatics. Languages with n ≥ 10.

Fig S16. Mean RQI by language and CI presence, bioinformatics. Means ± SE.

Fig S17. Mean RQI by language and testing presence, bioinformatics. Means ± SE.

13. Repository Age

Repositories are grouped into age cohorts; mean RQI is compared by CI and testing.

13.1 Age and CI

Fig 3A (main text). Mean RQI by age cohort and CI status. Age is years from first commit to data collection (March 2026); means ± SE.

13.2 Age and Testing

Fig 3B (main text). Mean RQI by age cohort and testing status. Means ± SE.

13.3 Age, CI and Testing (Combined)

Fig 3 (main text). Mean RQI by age cohort, by CI (a) and testing (b). Age is years from first commit to data collection (March 2026); means ± SE.

13.4 Cohort Gaps

RQI gaps (with minus without CI or tests) and Welch t-test p-values by cohort.

Table ST26. Mean RQI by age cohort, with and without CI (Welch t-test p values)

Age cohort No CI Has CI Gap p
0–2 years 3.08 3.47 0.39 <0.001
2–5 years 2.54 3.09 0.56 <0.001
5–10 years 2.39 2.90 0.51 <0.001
10+ years 2.42 3.17 0.75 <0.001

Table ST27. Mean RQI by age cohort, with and without tests (Welch t-test p values)

Age cohort No tests Has tests Gap p
0–2 years 3.15 3.38 0.22 <0.001
2–5 years 2.73 2.95 0.23 <0.001
5–10 years 2.41 2.74 0.32 <0.001
10+ years 2.46 2.84 0.38 <0.001

14. Citations and RQI

14.1 Median Citations, Published Repos

Fig 5 (main text). Median citations by RQI bin, published repositories. Repositories with at least one citation; color shows RQI (red = low, green = high). Brackets give exact p-values for significant Wilcoxon comparisons.

15. Publication Status

15.1 RQI Components

Fig 4 (main text), panel a. RQI and components by publication status, bioinformatics. Combined with panel b in Section 16.4.

16. Funding

16.1 Data

Table ST28. Funding data overview

Statistic Count
Repositories 1257
Grant funded 674
Not grant funded 583
With developer award and RQI 175

16.2 RQI Components (All Domains)

Fig S18. RQI and components by funding status, all domains.

16.3 Test Statistics (All Domains)

Table ST29. Wilcoxon rank-sum test of RQI by grant funding status (all domains)

Comparison n 1 n 2 Median 1 Median 2 W p Rank-biserial r
RQI (1 = Grant Funded, 2 = Not Grant Funded) 674 583 2.96 2.94 192724 0.559 0.019

16.4 RQI Components (Bioinformatics)

Fig 4 (main text), panel b. RQI and components by funding status, bioinformatics.

Grant sources: NIH, NSF, Other.

Table ST30. Wilcoxon rank-sum test of RQI by grant funding status (bioinformatics)

Comparison n 1 n 2 Median 1 Median 2 W p Rank-biserial r
RQI (1 = Grant Funded, 2 = Not Grant Funded) 473 238 2.91 2.71 61561.5 0.041 -0.094

Fig 4 (main text). RQI and components by publication (a) and grant funding (b), bioinformatics. Violins show RQI and its components; diamonds mark medians; Wilcoxon results annotated.

17. Citation Models

17.1 Transformations

Fig S19. Citation transformation diagnostics. Raw, log, and square-root histograms, residual plots, and Box–Cox profile.

Table ST31. Citation transformations: model R-squared and Box-Cox lambda

R-squared (log) R-squared (sqrt) R-squared (raw) Box-Cox lambda
0.019 0.013 0.001 -0.141

17.2 Citations vs RQI

Fig S20. Citations vs RQI, published repositories. OLS fit of log(citations + 1) with 95% CI.

Table ST32. Linear model of log(citations + 1) on RQI (published repositories)

Term Estimate Std. Error t p CI lower CI upper
Intercept 2.853 0.215 13.279 <0.001 2.432 3.275
RQI 0.313 0.070 4.499 <0.001 0.177 0.450

Table ST33. Model fit (log(citations + 1) on RQI)

n R-squared Adj. R-squared Residual SE Residual df F F df p
1037 0.019 0.018 1.737 1035 20.241 1, 1035 <0.001