IDEAL LEVEL OF MEASUREMENT

CENTRAL TENDENCY
VARIANCE
UNIVARITE INFERENCE
RELATIONSHIP OR ASSOCIATION
BIVARIATE INFERENCE
Nominal


Mode


IQV 


C.I. of Prop** One-sample t-test of Prop**


Symm. & Asym. lambda** Phi**


t-test of proportions**
Chi Square**


Ordinal


Median


IQV


Gamma, Gamma12.3  Tau, Tau12.3 


Gamma Test of Sig.(Var.#3 = NOM)**  Tau Test of Sig.** 


Interval


Mean


Range, IQR
Standard deviation


C.I. of Means** One-sample t-test of Means** 


r**, r12.3** 


r Test of Sig.**


int. by Nom.-


Eta 


t-test of means**  F-test** 


**The statistic has assumptions. For multiple choice "why or why not..." questions, ask:

  1. Can the statistic answer the question?
  2. Are the assumptions of the statistic met?
  3. Is the level of measurement OK for the statistic? Remember, although data can be wasted, the reverse is not true. A statistic constructed for a higher level of measurement cannot be used on a lower level of measurement.
  4. LEVEL OF MEASUREMENT

LEVEL OF STATISTIC


NOMINAL


ORDINAL


INTERVAL


NOMINAL


IDEAL


OK,    BUT 


WASTES


ORDINAL


LOGICAL


IDEAL


DATA


INTERVAL


   ERRORS


MADE


IDEAL


Hints:

  1. Be sure to ask the above three questions in the order given.
  2. If the answer to any question is "no," then do not ask the next question; instead, look for that "no" answer in the multiple choice answers.
  3. If the answer to any question is "yes," then go on to the next question.
  1. table of contents
  2. To return to the previous url use the browser's back button.