20.08.2013 Views

Suppes and Zinnes - basic measurement theory.pdf - Ted Sider

Suppes and Zinnes - basic measurement theory.pdf - Ted Sider

Suppes and Zinnes - basic measurement theory.pdf - Ted Sider

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

BASIC MEASUREMENT THEORY<br />

Theorem 34 (Uniqueness Theorem). If the successive interval system<br />

il3 = (B, k,f) is an A.M. system, then the derived scale (iB, R, !") is an<br />

interval scale in both the narrow <strong>and</strong> wide senses.<br />

Adams <strong>and</strong> Messick also note that if the values of f.i-a <strong>and</strong> ua are fixed<br />

for one stimulus or, alternatively, if one of the normal cumulatives associated<br />

with the stimuli is fixed, then the derived numerical assignment<br />

p is uniquely determined.<br />

For completeness we should point out that the numbers t 0 , tb ... , tk<br />

can also be thought of as the values of a derived numerical assignment for<br />

the category boundaries; that is, separate representation <strong>and</strong> uniqueness<br />

theorems can be established for the boundaries as well as for the stimuli,<br />

but since the theorems for the boundaries are similar to those just described<br />

for the stimuli we shall not pursue these details further.<br />

5. THE PROBLEM OF MEANINGFULNESS<br />

Closely related to the two fundamental issues of <strong>measurement</strong>-the<br />

representation <strong>and</strong> uniqueness problems-is a third problem which we<br />

shall term the meaningfulness problem. Although this problem is not<br />

central to a <strong>theory</strong> of <strong>measurement</strong>, it is often involved with various<br />

aspects of how <strong>measurement</strong>s may be used, <strong>and</strong> as such it has engendered<br />

considerable controversy. It therefore merits some discussion here,<br />

although necessarily our treatment will be curtailed.<br />

5.1 Definition<br />

To begin with, it will be well to illustrate the <strong>basic</strong> source of the meaningfulness<br />

problem with two of the statements, (4) <strong>and</strong> (5), given previously<br />

in Sec. 1.2. For convenience they are repeated here.<br />

4. The ratio of the maximum temperature today (tJ to the maximum<br />

temperature yesterday (tn_ 1 ) is 1.10.<br />

5. The ratio of the difference between today's <strong>and</strong> yesterday's maximum<br />

temperature Ctn <strong>and</strong> tn-J to the difference between today's <strong>and</strong><br />

tomorrow's maximum temperature Ctn <strong>and</strong> tn+1) will be 0.95.<br />

In Sec. 1.2 statement 4 was dismissed for having no clear empirical<br />

meaning <strong>and</strong> (5), on the other h<strong>and</strong>, was said to be acceptable. Here we<br />

wish to be completely explicit regarding the basis of this difference. Note<br />

first that both statements are similar in at least one important respect:<br />

neither statement specifies which numerical assignment from the set of<br />

THE PROBLEM OF MEANINGFULNESS<br />

admissible numerical assignments is to be used to determine the validity<br />

or truth of the statement. Are the temperature <strong>measurement</strong>s to be made<br />

on the centigrade, Fahrenheit, or possibly the Kelvin scale? This question<br />

is not answered by either statement.<br />

The distinguishing feature of ( 4) is that this ambiguity in the statement<br />

is critical. As an example, suppose by using the Fahrenheit scale we found<br />

that tn = 110 <strong>and</strong> 1n_ 1 = 100. We would then conclude that (4) was a<br />

true statement, the ratio of tn to tn_ 1 being 1.10. Note, however, that if<br />

the temperature <strong>measurement</strong>s had been made on the centigrade scale<br />

we would have come to the opposite conclusion, since then we would<br />

have found that tn = 43.3 <strong>and</strong> tn-l = 37.8 <strong>and</strong> a ratio equal to 1.15<br />

rather than 1.10. In the case of (4) the selection of a particular numerical<br />

assignment influences the conclusions we come to concerning the truth of<br />

the statement.<br />

In contrast, consider statement 5. The choice of a specific numerical<br />

assignment is not critical. To illustrate this point assume that when we<br />

measure temperature on a Fahrenheit scale we find the following readings:<br />

tn_ 1 = 60, tn = 79.0, <strong>and</strong> tn+ 1 = 99.0. Then the ratio described in (5) is<br />

equal to<br />

60- 79 = -19 = 0.95<br />

79- 99 -20<br />

which would indicate that (5) was valid. Now, if, instead, the temperature<br />

<strong>measurement</strong>s had been made on the centigrade scale, we would have<br />

found that tn_ 1 = 15.56, tn = 26.11, <strong>and</strong> tn+ 1 = 37.20. But, since the<br />

ratio of these numbers is<br />

15.56- 26.11 = -10.55 = 0.95,<br />

26.11- 37.20 -11.09<br />

we would also have come to the conclusion that (5) was true. In fact, it is<br />

easy to verify that if we restrict ourselves to numerical assignments which<br />

are linearly related to each other then we will always arrive at the same<br />

conclusion concerning the validity of (5). If the statement is true for one<br />

numerical assignment, then it will be true for all. And, furthermore, if<br />

it is untrue for one numerical assignment, then it will be untrue for all.<br />

For this reason we can say that the selection of a specific numerical<br />

assignment to test the statement is not critical.<br />

The absence of units in (4) <strong>and</strong> (5) is deliberate because the determination<br />

of units <strong>and</strong> an appreciation of their empirical signiHcance comes after,<br />

not before, the investigation of questions of meaningfulness. The character<br />

of (4) well illustrates this point. If the Fahrenheit scale is specified in (4),<br />

the result is an empirical statement that is unambiguously true or false, but

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!