The objections stopped and so did the objectors

# Self-authored for the EML case corpus (no external origin). A design review
# series records fewer objections every quarter. Where the objections went and
# where the objectors went are computed separately below.
#
# The review process improved and the improvements were real. Templates were
# added, proposals arrive with a written problem statement, and the authors
# have got much better at pre-socialising a design before the meeting. Any of
# those would reduce objections on its own, and all three happened.
#
# An objection is raised by a person who is present, and the set of people
# present is not constant. A falling objection rate is consistent with better
# designs and it is equally consistent with a smaller set of people willing to
# object, and the two are distinguished by counting attendance rather than by
# counting objections.
#
# Both are counted below.

# [quarter, proposals, objections raised, invited, attended, distinct people who objected]
[["Q1", 18, 41, 22, 19, 11], ["Q2", 20, 33, 22, 17, 9], ["Q3", 19, 22, 23, 14, 6], ["Q4", 21, 14, 24, 12, 4], ["Q5", 22, 9, 25, 11, 3], ["Q6", 20, 5, 26, 9, 2]] => quarters

len(quarters) => n

"quarter   proposals   objections   invited   attended   distinct objectors" ^0
for q in quarters:
    "  " + q[0] + "        " + str(q[1]) + "          " + str(q[2]) + "           " + str(q[3]) + "        " + str(q[4]) + "         " + str(q[5]) ^0
"" ^0

quarters[0] => first
quarters[n - 1] => last
"objections per proposal" ^0
for q in quarters:
    "  " + q[0] + " : " + str(int(q[2] * 100 / q[1])) + " per 100 proposals" ^0
"  " + first[0] + " to " + last[0] + " : " + str(int(first[2] * 100 / first[1])) + " -> " + str(int(last[2] * 100 / last[1])) ^0
"  a fall of " + str(int((first[2] * 100 / first[1] - last[2] * 100 / last[1]) * 100 / (first[2] * 100 / first[1]))) + "%" ^0
"" ^0

# ---- who was in the room ----

"attendance against invitation" ^0
for q in quarters:
    "  " + q[0] + " : " + str(q[4]) + " of " + str(q[3]) + " invited, " + str(int(q[4] * 100 / q[3])) + "%" ^0
"  " + first[0] + " : " + str(int(first[4] * 100 / first[3])) + "%, " + last[0] + " : " + str(int(last[4] * 100 / last[3])) + "%" ^0
if last[4] < first[4]:
    "  the invitation list grew by " + str(last[3] - first[3]) + " and the room shrank by " + str(first[4] - last[4]) ^0
"" ^0

# ---- objections per person who was actually there ----

"objections per attendee, which holds the room size constant" ^0
for q in quarters:
    "  " + q[0] + " : " + str(int(q[2] * 100 / q[4])) + " per 100 attendees" ^0
int(first[2] * 100 / first[4]) => rate_first
int(last[2] * 100 / last[4]) => rate_last
"  " + first[0] + " : " + str(rate_first) + ", " + last[0] + " : " + str(rate_last) ^0
"  fall : " + str(int((rate_first - rate_last) * 100 / rate_first)) + "%" ^0
"  the per-proposal fall was " + str(int((first[2] * 100 / first[1] - last[2] * 100 / last[1]) * 100 / (first[2] * 100 / first[1]))) + "%, so most of it survives the correction" ^0
"  and some of it does not" ^0
"" ^0

# ---- the objectors themselves ----

"how many different people raised anything" ^0
for q in quarters:
    "  " + q[0] + " : " + str(q[5]) + " people, " + str(int(q[5] * 100 / q[4])) + "% of the room" ^0
"  " + first[0] + " : " + str(first[5]) + " of " + str(first[4]) + " attendees" ^0
"  " + last[0] + " : " + str(last[5]) + " of " + str(last[4]) + " attendees" ^0
"  the number of people who object at all has fallen by " + str(int((first[5] - last[5]) * 100 / first[5])) + "%" ^0
"" ^0

# ---- where the objectors went ----

# [person, quarters they objected in, still attending at Q6, where they are now]
[["a", 6, "yes", "still here"], ["b", 5, "no", "moved to another team"], ["c", 4, "no", "left the company"], ["d", 4, "no", "stopped attending"], ["e", 3, "no", "stopped attending"], ["f", 3, "no", "moved to another team"], ["g", 2, "yes", "still here"], ["h", 2, "no", "left the company"], ["i", 1, "no", "stopped attending"], ["j", 1, "no", "stopped attending"], ["k", 1, "no", "stopped attending"]] => objectors

len(objectors) => m
0 => remaining
0 => quiet
0 => gone
for p in objectors:
    if p[2] == "yes":
        remaining + 1 => remaining
    if p[3] == "stopped attending":
        quiet + 1 => quiet
    if p[3] == "left the company":
        gone + 1 => gone
"the " + str(m) + " people who ever raised an objection in this series" ^0
"  still attending at " + last[0] + " : " + str(remaining) ^0
"  stopped attending, still at the company : " + str(quiet) ^0
"  left the company : " + str(gone) ^0
"  moved to another team : " + str(m - remaining - quiet - gone) ^0
"  so " + str(m - remaining) + " of " + str(m) + " are not in the room any more" ^0
"" ^0

# ---- the two explanations, and what separates them ----

"what a falling objection count is consistent with" ^0
"  the designs got better       : consistent" ^0
"  the room got smaller         : consistent" ^0
"  the people who object left   : consistent" ^0
"  all three are happening here, and the objection count alone cannot" ^0
"  apportion them" ^0
"" ^0

"what would separate them" ^0
"  ask a departed objector to review a current proposal" ^0
"  proposals reviewed that way : 0" ^0
"  so the measurement that would distinguish the explanations has not been" ^0
"  taken, and the one that has been taken cannot" ^0
"" ^0

# ---- the control: a review series with stable attendance ----
#
# Where the same people keep attending, a falling objection rate is a fact
# about the proposals rather than about the roster.

[["Q1", 12, 20, 12, 12, 7], ["Q6", 13, 8, 12, 12, 6]] => stable
"control - a second series where attendance did not move" ^0
for s in stable:
    "  " + s[0] + " : " + str(s[2]) + " objections, " + str(s[4]) + " of " + str(s[3]) + " attending, " + str(s[5]) + " objectors" ^0
stable[0] => s_first
stable[1] => s_last
if s_first[4] == s_last[4]:
    "  the room is the same size and " + str(s_last[5]) + " of the " + str(s_first[5]) + " objectors are still in it" ^0
    "  objections fell " + str(int((s_first[2] - s_last[2]) * 100 / s_first[2])) + "% with the roster held constant, which is the version of" ^0
    "  this number that means what it appears to mean" ^0
"" ^0

"The templates and the pre-socialising are real improvements and objections" ^0
"really are lower. An objection needs somebody present who will raise it, so" ^0
"the count also measures the roster, and here the roster changed." ^0

quarters = [["Q1", 18, 41, 22, 19, 11], ["Q2", 20, 33, 22, 17, 9], ["Q3", 19, 22, 23, 14, 6], ["Q4", 21, 14, 24, 12, 4], ["Q5", 22, 9, 25, 11, 3], ["Q6", 20, 5, 26, 9, 2]]
n = len(quarters)
print("quarter   proposals   objections   invited   attended   distinct objectors")
for q in quarters:
    print("  " + q[0] + "        " + str(q[1]) + "          " + str(q[2]) + "           " + str(q[3]) + "        " + str(q[4]) + "         " + str(q[5]))
print("")
first = quarters[0]
last = quarters[n - 1]
print("objections per proposal")
for q in quarters:
    print("  " + q[0] + " : " + str(int(q[2] * 100 / q[1])) + " per 100 proposals")
print("  " + first[0] + " to " + last[0] + " : " + str(int(first[2] * 100 / first[1])) + " -> " + str(int(last[2] * 100 / last[1])))
print("  a fall of " + str(int((first[2] * 100 / first[1] - last[2] * 100 / last[1]) * 100 / (first[2] * 100 / first[1]))) + "%")
print("")
print("attendance against invitation")
for q in quarters:
    print("  " + q[0] + " : " + str(q[4]) + " of " + str(q[3]) + " invited, " + str(int(q[4] * 100 / q[3])) + "%")
print("  " + first[0] + " : " + str(int(first[4] * 100 / first[3])) + "%, " + last[0] + " : " + str(int(last[4] * 100 / last[3])) + "%")
if last[4] < first[4]:
    print("  the invitation list grew by " + str(last[3] - first[3]) + " and the room shrank by " + str(first[4] - last[4]))
print("")
print("objections per attendee, which holds the room size constant")
for q in quarters:
    print("  " + q[0] + " : " + str(int(q[2] * 100 / q[4])) + " per 100 attendees")
rate_first = int(first[2] * 100 / first[4])
rate_last = int(last[2] * 100 / last[4])
print("  " + first[0] + " : " + str(rate_first) + ", " + last[0] + " : " + str(rate_last))
print("  fall : " + str(int((rate_first - rate_last) * 100 / rate_first)) + "%")
print("  the per-proposal fall was " + str(int((first[2] * 100 / first[1] - last[2] * 100 / last[1]) * 100 / (first[2] * 100 / first[1]))) + "%, so most of it survives the correction")
print("  and some of it does not")
print("")
print("how many different people raised anything")
for q in quarters:
    print("  " + q[0] + " : " + str(q[5]) + " people, " + str(int(q[5] * 100 / q[4])) + "% of the room")
print("  " + first[0] + " : " + str(first[5]) + " of " + str(first[4]) + " attendees")
print("  " + last[0] + " : " + str(last[5]) + " of " + str(last[4]) + " attendees")
print("  the number of people who object at all has fallen by " + str(int((first[5] - last[5]) * 100 / first[5])) + "%")
print("")
objectors = [["a", 6, "yes", "still here"], ["b", 5, "no", "moved to another team"], ["c", 4, "no", "left the company"], ["d", 4, "no", "stopped attending"], ["e", 3, "no", "stopped attending"], ["f", 3, "no", "moved to another team"], ["g", 2, "yes", "still here"], ["h", 2, "no", "left the company"], ["i", 1, "no", "stopped attending"], ["j", 1, "no", "stopped attending"], ["k", 1, "no", "stopped attending"]]
m = len(objectors)
remaining = 0
quiet = 0
gone = 0
for p in objectors:
    if p[2] == "yes":
        remaining = remaining + 1
    if p[3] == "stopped attending":
        quiet = quiet + 1
    if p[3] == "left the company":
        gone = gone + 1
print("the " + str(m) + " people who ever raised an objection in this series")
print("  still attending at " + last[0] + " : " + str(remaining))
print("  stopped attending, still at the company : " + str(quiet))
print("  left the company : " + str(gone))
print("  moved to another team : " + str(m - remaining - quiet - gone))
print("  so " + str(m - remaining) + " of " + str(m) + " are not in the room any more")
print("")
print("what a falling objection count is consistent with")
print("  the designs got better       : consistent")
print("  the room got smaller         : consistent")
print("  the people who object left   : consistent")
print("  all three are happening here, and the objection count alone cannot")
print("  apportion them")
print("")
print("what would separate them")
print("  ask a departed objector to review a current proposal")
print("  proposals reviewed that way : 0")
print("  so the measurement that would distinguish the explanations has not been")
print("  taken, and the one that has been taken cannot")
print("")
stable = [["Q1", 12, 20, 12, 12, 7], ["Q6", 13, 8, 12, 12, 6]]
print("control - a second series where attendance did not move")
for s in stable:
    print("  " + s[0] + " : " + str(s[2]) + " objections, " + str(s[4]) + " of " + str(s[3]) + " attending, " + str(s[5]) + " objectors")
s_first = stable[0]
s_last = stable[1]
if s_first[4] == s_last[4]:
    print("  the room is the same size and " + str(s_last[5]) + " of the " + str(s_first[5]) + " objectors are still in it")
    print("  objections fell " + str(int((s_first[2] - s_last[2]) * 100 / s_first[2])) + "% with the roster held constant, which is the version of")
    print("  this number that means what it appears to mean")
print("")
print("The templates and the pre-socialising are real improvements and objections")
print("really are lower. An objection needs somebody present who will raise it, so")
print("the count also measures the roster, and here the roster changed.")

quarter   proposals   objections   invited   attended   distinct objectors
  Q1        18          41           22        19         11
  Q2        20          33           22        17         9
  Q3        19          22           23        14         6
  Q4        21          14           24        12         4
  Q5        22          9           25        11         3
  Q6        20          5           26        9         2

objections per proposal
  Q1 : 227 per 100 proposals
  Q2 : 165 per 100 proposals
  Q3 : 115 per 100 proposals
  Q4 : 66 per 100 proposals
  Q5 : 40 per 100 proposals
  Q6 : 25 per 100 proposals
  Q1 to Q6 : 227 -> 25
  a fall of 89%

attendance against invitation
  Q1 : 19 of 22 invited, 86%
  Q2 : 17 of 22 invited, 77%
  Q3 : 14 of 23 invited, 60%
  Q4 : 12 of 24 invited, 50%
  Q5 : 11 of 25 invited, 44%
  Q6 : 9 of 26 invited, 34%
  Q1 : 86%, Q6 : 34%
  the invitation list grew by 4 and the room shrank by 10

objections per attendee, which holds the room size constant
  Q1 : 215 per 100 attendees
  Q2 : 194 per 100 attendees
  Q3 : 157 per 100 attendees
  Q4 : 116 per 100 attendees
  Q5 : 81 per 100 attendees
  Q6 : 55 per 100 attendees
  Q1 : 215, Q6 : 55
  fall : 74%
  the per-proposal fall was 89%, so most of it survives the correction
  and some of it does not

how many different people raised anything
  Q1 : 11 people, 57% of the room
  Q2 : 9 people, 52% of the room
  Q3 : 6 people, 42% of the room
  Q4 : 4 people, 33% of the room
  Q5 : 3 people, 27% of the room
  Q6 : 2 people, 22% of the room
  Q1 : 11 of 19 attendees
  Q6 : 2 of 9 attendees
  the number of people who object at all has fallen by 81%

the 11 people who ever raised an objection in this series
  still attending at Q6 : 2
  stopped attending, still at the company : 5
  left the company : 2
  moved to another team : 2
  so 9 of 11 are not in the room any more

what a falling objection count is consistent with
  the designs got better       : consistent
  the room got smaller         : consistent
  the people who object left   : consistent
  all three are happening here, and the objection count alone cannot
  apportion them

what would separate them
  ask a departed objector to review a current proposal
  proposals reviewed that way : 0
  so the measurement that would distinguish the explanations has not been
  taken, and the one that has been taken cannot

control - a second series where attendance did not move
  Q1 : 20 objections, 12 of 12 attending, 7 objectors
  Q6 : 8 objections, 12 of 12 attending, 6 objectors
  the room is the same size and 6 of the 7 objectors are still in it
  objections fell 60% with the roster held constant, which is the version of
  this number that means what it appears to mean

The templates and the pre-socialising are real improvements and objections
really are lower. An objection needs somebody present who will raise it, so
the count also measures the roster, and here the roster changed.

Trace event types