Group subjects by the study's baseline BMI quartiles

1library(yamaa)23yamaa_spec("spec.yaml")$run()$output

1import yamaa23yamaa.spec("spec.yaml").run().output

1# Reference solution for the yamaa benchmark adam-adsl-bmi-quartile-group (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)910adsl_in <- read_csv(11  "/app/input/adsl.csv",12  col_types = cols(BMIBL = col_double(), .default = col_character())13)1415# R's default quantile definition is type 7.16quartile <- function(x, p) {17  quantile(x, p, type = 7, na.rm = TRUE, names = FALSE)18}1920adsl <- adsl_in |>21  group_by(STUDYID) |>22  mutate(23    BMIBLQ1 = quartile(BMIBL, 0.25),24    BMIBLMED = quartile(BMIBL, 0.5),25    BMIBLQ3 = quartile(BMIBL, 0.75)26  ) |>27  ungroup() |>28  mutate(29    BMIBLGR1 = case_when(30      is.na(BMIBL) ~ NA_character_,31      BMIBL <= BMIBLQ1 ~ "Quartile 1",32      BMIBL <= BMIBLMED ~ "Quartile 2",33      BMIBL <= BMIBLQ3 ~ "Quartile 3",34      .default = "Quartile 4"35    )36  ) |>37  select(STUDYID, USUBJID, BMIBL, BMIBLQ1, BMIBLMED, BMIBLQ3, BMIBLGR1)3839dir.create("/app/output", showWarnings = FALSE)40write_csv(adsl, "/app/output/adsl.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adsl-bmi-quartile-group (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67import csv8import math910INPUT_CSV = "/app/input/adsl.csv"11OUTPUT_CSV = "/app/output/adsl.csv"121314def quantile(values, p):15    """R's default quantile (type 7) of the values at proportion p."""16    xs = sorted(values)17    h = (len(xs) - 1) * p + 118    j = math.floor(h)19    g = h - j20    if g == 0 or j == len(xs):21        return xs[j - 1]22    return xs[j - 1] + g * (xs[j] - xs[j - 1])232425def group(bmi, q1, med, q3):26    if bmi is None:27        return ""28    if bmi <= q1:29        return "Quartile 1"30    if bmi <= med:31        return "Quartile 2"32    if bmi <= q3:33        return "Quartile 3"34    return "Quartile 4"353637def cell(value):38    return "" if value is None else repr(value)394041def main():42    with open(INPUT_CSV, newline="") as f:43        rows = list(csv.DictReader(f))44    for row in rows:45        text = row["BMIBL"].strip()46        row["value"] = float(text) if text else None4748    cutpoints = {}49    for study in {row["STUDYID"] for row in rows}:50        values = [r["value"] for r in rows if r["STUDYID"] == study and r["value"] is not None]51        cutpoints[study] = tuple(quantile(values, p) for p in (0.25, 0.5, 0.75))5253    fieldnames = ["STUDYID", "USUBJID", "BMIBL", "BMIBLQ1", "BMIBLMED", "BMIBLQ3", "BMIBLGR1"]54    with open(OUTPUT_CSV, "w", newline="") as f:55        writer = csv.DictWriter(f, fieldnames=fieldnames)56        writer.writeheader()57        for row in rows:58            q1, med, q3 = cutpoints[row["STUDYID"]]59            writer.writerow(60                {61                    "STUDYID": row["STUDYID"],62                    "USUBJID": row["USUBJID"],63                    "BMIBL": cell(row["value"]),64                    "BMIBLQ1": cell(q1),65                    "BMIBLMED": cell(med),66                    "BMIBLQ3": cell(q3),67                    "BMIBLGR1": group(row["value"], q1, med, q3),68                }69            )707172if __name__ == "__main__":73    main()

1/* opensas reference solution for the yamaa benchmark adam-adsl-bmi-quartile-group.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas.4   The quartiles follow R's default definition (type 7), which none of the5   percentile definitions of the procedures gives, so a DATA step computes6   them from the sorted values. */78data adsl;9  length STUDYID USUBJID $40;10  infile "/app/input/adsl.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;11  input STUDYID : $char40. USUBJID : $char40. BMIBL;12run;1314/* The baseline values of each study in ascending order, numbered from 1. */15proc sort data=adsl(where=(not missing(BMIBL))) out=sorted;16  by STUDYID BMIBL;17run;1819data ranked;20  set sorted;21  by STUDYID;22  if first.STUDYID then RANK = 0;23  RANK + 1;24  keep STUDYID RANK BMIBL;25run;2627proc sql;28  create table counts as29    select STUDYID, count(*) as N from ranked group by STUDYID;30quit;3132/* Type 7: the quantile at proportion P sits at position H = (N - 1) * P + 1,33   between the values ranked J = floor(H) and K = J + 1, at fraction G. */34data positions;35  set counts;36  STAT = 1; P = 0.25; H = (N - 1) * P + 1; J = floor(H); G = H - J;37  K = min(J + 1, N); output;38  STAT = 2; P = 0.5; H = (N - 1) * P + 1; J = floor(H); G = H - J;39  K = min(J + 1, N); output;40  STAT = 3; P = 0.75; H = (N - 1) * P + 1; J = floor(H); G = H - J;41  K = min(J + 1, N); output;42run;4344proc sql;45  create table quartiles as46    select p.STUDYID, p.STAT, lo.BMIBL + p.G * (hi.BMIBL - lo.BMIBL) as Q47    from positions as p48      inner join ranked as lo on lo.STUDYID = p.STUDYID and lo.RANK = p.J49      inner join ranked as hi on hi.STUDYID = p.STUDYID and hi.RANK = p.K;50  create table cutpoints as51    select STUDYID,52           max(case when STAT = 1 then Q else . end) as BMIBLQ1,53           max(case when STAT = 2 then Q else . end) as BMIBLMED,54           max(case when STAT = 3 then Q else . end) as BMIBLQ355    from quartiles group by STUDYID;56  create table joined as57    select a.STUDYID, a.USUBJID, a.BMIBL, c.BMIBLQ1, c.BMIBLMED, c.BMIBLQ358    from adsl as a left join cutpoints as c on c.STUDYID = a.STUDYID;59quit;6061data grouped;62  set joined;63  length BMIBLGR1 $10;64  if missing(BMIBL) then BMIBLGR1 = '';65  else if BMIBL <= BMIBLQ1 then BMIBLGR1 = 'Quartile 1';66  else if BMIBL <= BMIBLMED then BMIBLGR1 = 'Quartile 2';67  else if BMIBL <= BMIBLQ3 then BMIBLGR1 = 'Quartile 3';68  else BMIBLGR1 = 'Quartile 4';69run;7071proc sql;72  create table final as73    select STUDYID, USUBJID, BMIBL, BMIBLQ1, BMIBLMED, BMIBLQ3, BMIBLGR174    from grouped order by STUDYID, USUBJID;75quit;76proc export data=final outfile="/app/output/adsl.csv" dbms=csv replace; run;