Measurement, Meaning

and Morality
John Michael Linacre, Ph.D.
University of Sydney, Australia

I. Measurement
Science Measurement Society ???
Treaty of the Meter - 1875

I. Measurement
Science Measurement Society ???
Treaty of the Meter - 1875
Society Measurement Science !!!
Babylonian land survey - 1500 BCE

I. Measurement
Science Measurement Society ???
Treaty of the Meter - 1875
Society Measurement Science !!!
Babylonian land survey - 1500 BCE
Measurement = Size or Quantity

I. Measurement
Science Measurement Society ???
Treaty of the Meter - 1875
Society Measurement Science !!!
Babylonian land survey - 1500 BCE
Measurement = Size or Quantity
Quantity = Numerable Amount

I. Measurement
Science Measurement Society ???
Treaty of the Meter - 1875
Society Measurement Science !!!
Babylonian land survey - 1500 BCE
Measurement = Size or Quantity
Quantity = Numerable Amount

Measurement is the imposition of

the rules of arithmetic on the world
around us.

One Unit Extra

One more unit means the same amount extra,
no matter how much we already have.

One Unit Extra

One more unit means the same amount extra,
no matter how much we already have.
One more Orange = One more unit of Juice ???

One Unit Extra

One more unit means the same amount extra,
no matter how much we already have.
One more Orange = One more unit of Juice ???

One more Orange = 59 cc. 177 cc. of Juice

Texas Department of Agriculture

One Unit Extra

One more unit means the same amount extra,
no matter how much we already have.
One more Orange = One more unit of Juice ???

One more Orange = 59 cc. 177 cc. of Juice

Texas Department of Agriculture

So, we trade Oranges by

abstract arithmetical units of volume or weight!

Measures: Numbers arent enough .

Earthquakes

Measures: Numbers arent enough .

Earthquakes

???

Measures: Clarity requires additivity

Earthquakes

!!!

Educational and Psychological Additive Units?

Measurement, in any true sense, is
impossible in psychology, but their opinion
might change if new facts were established
Final Report BAAS, 1940 ???

Educational and Psychological Additive Units?

Measurement, in any true sense, is
impossible in psychology, but their opinion
might change if new facts were established
Final Report BAAS, 1940 ???

the stars .... we would never by any means

investigate their chemical composition
Auguste Comte, 1842 ???

Educational and Psychological Additive Units?

Measurement, in any true sense, is
impossible in psychology, but their opinion
might change if new facts were established
Final Report BAAS, 1940 ???

the stars .... we would never by any means

investigate their chemical composition
Auguste Comte, 1842 ???
1859, Gustav Kirchoff - spectral analysis of Sun !!!

Old Facts of Measurement: Concatenation

Rod(A)
Rod(B)
= Rod(A+B)

N. Campbell: additivity requires rules of concatenation:

The length rule is: Place rods end-to-end !!!

Old Facts of Measurement: Concatenation

Rod(A)
Rod(B)
= Rod(A+B)

N. Campbell: additivity requires rules of concatenation:

The length rule is: Place rods end-to-end !!!
We need the psychological concatenation rule for:
Outcome(Bni)
Outcome(Bmi)

Person n
Person m

=Outcome(Bni+Bmi)

Item i

???

Conjecturing a rule based on probability ..

Let Pni be the probability of success of
person n on item i, inferred from data:
0 Pni 1

Conjecturing a rule based on probability ..

Let Pni be the probability of success of
person n on item i, inferred from data:
0 Pni 1
Commensurate with an infinite latent variable:
0 Pni / (1-Pni)

Conjecturing a rule based on probability ..

Let Pni be the probability of success of
person n on item i, inferred from data:
0 Pni 1
Commensurate with an infinite latent variable:
0 Pni / (1-Pni)
- log(Pni / (1-Pni))

Conjecturing a rule based on probability ..

Let Pni be the probability of success of
person n on item i, inferred from data:
0 Pni 1
Commensurate with an infinite latent variable:
0 Pni / (1-Pni)
- log(Pni / (1-Pni))
Suppose: Outcome(Bni) = log(Pni / (1-Pni))

Outcome(Bmi) = log(Pmi / (1-Pmi))

Then: Outcome(Bni) + Outcome(Bmi) = ..

Revising a rule based on probability ..

Outcome(Bni) + Outcome(Bmi) =
log(Pni / (1-Pni)) + log(Pmi / (1-Pmi))
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )

Revising a rule based on probability ..

Outcome(Bni) + Outcome(Bmi) =
log(Pni / (1-Pni)) + log(Pmi / (1-Pmi))
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )
So, a revised concatenation rule: Let Outcome be
the log-odds of coincident scored observations

Revising a rule based on probability ..

Outcome(Bni) + Outcome(Bmi) =
log(Pni / (1-Pni)) + log(Pmi / (1-Pmi))
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )
So, a revised concatenation rule: Let Outcome be
the log-odds of coincident scored observations
Due to self-coincidence, Outcome(Bni) is unchanged .

Revising a rule based on probability ..

Outcome(Bni) + Outcome(Bmi) =
log(Pni / (1-Pni)) + log(Pmi / (1-Pmi))
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )
So, a revised concatenation rule: Let Outcome be
the log-odds of coincident scored observations
Due to self-coincidence, Outcome(Bni) is unchanged .

Outcome (Bni+Bmi) = log( Probability of coinciding

on success / Probability of coinciding on failure )
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )

Revising a rule based on probability ..

Outcome(Bni) + Outcome(Bmi) =
log(Pni / (1-Pni)) + log(Pmi / (1-Pmi))
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )
So, a revised concatenation rule: Let Outcome be
the log-odds of coincident scored observations
Due to self-coincidence, Outcome(Bni) is unchanged .

Outcome (Bni+Bmi) = log( Probability of coinciding

on success / Probability of coinciding on failure )
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )
= Outcome(Bni) + Outcome(Bmi) !!!

Revising a rule based on probability ..

Outcome(Bni) + Outcome(Bmi) =
log(Pni / (1-Pni)) + log(Pmi / (1-Pmi))
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )
So, a revised concatenation rule: Let Outcome be
the log-odds of coincident scored observations
Due to self-coincidence, Outcome(Bni) is unchanged .

Outcome (Bni+Bmi) = log( Probability of coinciding

on success / Probability of coinciding on failure )
= log( (Pni * Pmi) / ((1-Pni)*(1-Pmi)) )
= Outcome(Bni) + Outcome(Bmi) !!!
A concatenation rule for people taking test items!!

Old Facts of Measurement: Concatenation

Rod(A)
Rod(B)
= Rod(A+B)

N. Campbell: additivity requires rules of concatenation:

We have the psychological concatenation rule for:

Outcome(Bni)
Outcome(Bmi)

Person n
Person m

=Outcome(Bni+Bmi)

Item i

!!!

A Glorious New Fact !!!

The Final Report was not the last word .
There can be measurement in Education,
Psychology and the Social Sciences equally
as valid and rugged as in Physics!
But Social Science measurement requires
painstaking effort, exactly as in Physics

The simple dichotomous Rasch model ..

Outcome(Bni) = log(Pni / (1-Pni))

Define: Outcome(Bni) = Bn - Di
Bn - Di = log(Pni / (1-Pni))
which can be rewritten as the Rasch model:
!!!

The simple dichotomous Rasch model ..

Outcome(Bni) = log(Pni / (1-Pni))

Define: Outcome(Bni) = Bn - Di
Bn - Di = log(Pni / (1-Pni))
which can be rewritten as the Rasch model:
!!!
Finding: the Rasch model is a concatenation rule, a
new fact (Rasch, 1953), operationalizing true
measurement for education and psychology.

II. Measurement and Meaning

Mohammed ibn-Musa al-Khowarizmi (830 CE,
Algorithm)
Book: Hisab al-jabr w'al muqabala (Algebra)
Calculating by restoring and comparing

What we want to find out is the thing,

a value on a (latent) variable:
= shay (Arabic) heard as xay (Spanish) x (English)
So meaning is expressed as measures on an x-axis
representing amounts of the latent variable thing.

Mallinson
(2001)

Level of Difficulty

How demanding/fatiguing is the activity?

Less Demanding
(Fatiguing) Activity

Getting out
of bed

Bathe & dress

Yard work

Sports
activity

Equal Intervals

Most Fatigued
Person

More Demanding
(Fatiguing) Activity

Level of Fatigue
How fatigued is the person?

10

11

Least Fatigued
Person

WRAT3 Item Map and Absolute Scale

III. Measurement, Meaning and Morality

Measures must be:
ideal, but practical
rigorous, but accommodating
demanding, but forgiving
quantitative, but qualitative
forward-looking, but faithful to the past

fair, even-handed, honest

In Science:
The development of [physical] metrology
shows that the same principle is being fulfilled in
physics. The accord of ethical and physical
principles was first noted by Sir Arthur Eddington,
when in 1920 he chose the words from The Book of
Deuteronomy* as an epigraph to the chapter on
Weyls unified theory in his Space, Time and
Gravitation. ... We obviously live in the world
where the fundamental principles of ethics and
physics agree with each other. !!!
(Tomilin, 1999)

In Politics: the good

!!!
Chau conferred great gifts, and the good were
enriched. ... He carefully attended to the
weights and measures, and the good
government of the kingdom took its course.
(Confucius, The Analects, 20. ca. 500 BCE).

In Politics: the good

!!!
Chau conferred great gifts, and the good were
enriched. ... He carefully attended to the
weights and measures, and the good
government of the kingdom took its course.
(Confucius, The Analects, 20. ca. 500 BCE).

But what if . ???

In Politics: the bad

Sixty thousand measures of weight in France
before the Revolution of 1789. ... the falsification
of standards by the feudal land owners and the
distrust, justified or not, of the peasants. A common
demand was to unify weights and measures - not
to avoid paying feudal dues but to assure an
honest amount payable. The rallying cry: un roi,
une loi, un poids, et une mesure (one king, one
law, one weight, and one measure) was a slogan of
equality and centralization, one that the
Revolution furthered.
(Kennedy, 1989)

In Religion:

And O my people! give just measure and weight

The Quran, The Prophet Hud, 11:85, ca. 600 CE

In Religion:

And O my people! give just measure and weight

The Quran, The Prophet Hud, 11:85, ca. 600 CE

Surely this injunction applies equally to

educational, psychological and physical measures.

Measurement, meaning and morality

- honestly, they work together.

Measurement, Meaning
and Morality
John Michael Linacre, Ph.D.
University of Sydney, Australia