The instrument · at scale

The Field

What a sample would let the model say, at any size, on one finished record.

500respondents
sees r ≥ 0.2435 of 11 sized relationships seen
  1. 90Factor structurewhether nine indicators are one thing or several
  2. 217Normsevery reading against a measured mean, known to two points
  3. 300A held-out halfsix in ten discover, the rest confirm, nothing crosses
  4. 375Reliability, measureda 4-item scale’s alpha known to a twentieth
  5. 418The typical published effectr = 0.27, visible across all 820 pairs
  6. 681The pre-specified combinationseleven moderations and paths, corrected as a family
  7. 1,473Every three-way combination10,660 of them, corrected for all of them
  8. 3,023The horizonthe smallest published effect on the map, r = 0.10
n = 500300 discover · 200 confirmsees r ≥ 0.2435 of 11 sized seen · 36 unsized testable
One record · what spoke
Political Identification · 6
a leftist
Institutional Trust · 6
disaffected
Quote Resonance · 90
a populist
Social Dominance · 13
an egalitarian
Right-Wing Authoritarianism · 22
an individualist
Moral Foundations · 71
a humanist
Against the map
22
of 47 this record does not fit
29
a respondent with no structure at all
98%
of them reach 22 too
this record0 misses47

The map’s agreement scale cannot tell this record from a random one, so the 22 are the seed a population would test, never findings about the person.

Moved between sittings
Anonymous Reception
Named Reception
Misattributed Reception
Figure Sentiment
Continuity Reading
Liberty
Order
Regime Attribution
Bandwidth
Equality
The count that could be a finding
0
chances to be wrong about this person
47 relationships drawn · 0 carry a sourced correlation

A chance needs a sourced correlation, and none is sourced yet. That count moves with reading, not with respondents.

The other currency · what reading buys
0
chances to be wrong today · 0 of 47 sourced
2
chances once the 2 same-pair sizes are on their edges
2
look like misses on this record · 2 survive the correction

Each square is a drawn relationship; the filled ones carry a published size somewhere in the literature. Only a size measured on these two constructs can grade anybody, so the projection rests on 2 of them; the other 9 are broader or neighbouring findings, which say whether the design could see an effect that big and not what to expect of a person. Putting same-pair sizes on the edges is reading, not recruiting, and it is what makes the model falsifiable for one record before a second person answers.

What no sample buys
the design sees r ≥ 0.24380% of the spread kepteveryone differseveryone agrees.30.10

A true r of .30 on left–right self-placement, as a readership agrees more. 12 of the 47 drawn relationships end at self-placement; in a readership that opposes the same arrangement, a larger sample makes them more precise and no more visible.

The design in force
500
registered target · 300 discover, 200 confirm
r ≥ 0.243
what it sees at 820 pairs · the discovery half binds
3,023
respondents for the smallest sized relationship on the map, r = 0.10
  • A sourced size, at any n. Those come from papers.
  • A reliability better than the instrument’s own. More people measure it more precisely and cannot raise it.
  • Direction. Every relationship here is correlational at every n.