What are Variables

Classification of Variables and Characteristics of a Good Variable

What are Variables, a “good” variable, Construct validity,Variable Classification by statistical type,Tricky Variable Cases

Defining and Classifying Data Variables

    "The link from scientific concepts to data quantities."

    A key component of the design of experiments is operationalization, which is the formal procedure that links scientific concepts to data collection. Operationalizations define measures or variables which are quantities of interest or which serve as the practical substitutes for the concepts of interest.

What makes a “good” variable?

    Regardless of what we are trying to measure, the qualities that make a good measure of a scientific concept are high reliability, absence of bias, low cost, practicality, objectivity, high acceptance, and high concept validity. Reliability is essentially the inverse of the statistical concept of variance, and a rough equivalent is "consistency". Statisticians also use the word "precision". Bias refers to the difference between the measure and some “true” value. 

    A difference between an individual measurement and the true value is called an “error” (which implies the practical impossibility of perfect precision, rather than the making of mistakes). The bias is the average difference over many measurements. 

    Ideally the bias of a measurement process should be zero. For example, a measure of weight that is made with people wearing their street clothes and shoes has a positive bias equal to the average weight of the shoes and clothes across all subjects.

“Precision or reliability refers to the reproducibility of repeated measurements, while bias refers to how far the average of many measurements is from the true value.”

    All other things being equal, when two measures are available, we will choose the less expensive and easier to obtain (more practical) measures. Measures that have a greater degree of subjectivity are generally less preferable. 

    Although devising your own measures may improve upon existing measures, there may be a trade off with acceptability, resulting in reduced impact of your experiment on the field as a whole.

Construct validity is a key criterion for variable definition. Under ideal conditions, after completing your experiment you will be able to make a strong claim that changing your explanatory variable(s) in a certain way (eg, doubling the amplitude of a background hum) causes a corresponding change in your outcome (eg, score on an irritability scale). 

    But if you want to convert that to meaningful statements about the effects of auditory environmental disturbances on the psychological trait or construct called “irritability”, you must be able to argue that the scales have good construct validity for the traits, namely that the operationalization of background noise as an electronic hum has good construct validity for auditory environmental disturbances, and that your irritability scale really measures what people call irritability.

“Construct validity is the link from practical measurements to meaningful concepts.”

Classification by role

    There are two different independent systems of classification of variables that you must learn in order to understand the rest of this book. The first system is based on the role of the variable in the experiment and the analysis. The general terms used most frequently in this text are explanatory variables vs. outcome variables . 

    An experiment is designed to test the effects of some intervention on one or more measures, which are therefore designated as outcome variables. Much of this book deals with the most common type of experiment in which there is only a single outcome variable measured on each experimental unit (person, animal, factory, etc.) A synonym for outcome variable is dependent variable, often abbreviated DV.

    The second main role a variable may play is that of an explanatory variable. Explanatory variables include variables purposely manipulated in an experiment and variables that are not purposely manipulated, but are thought to possibly affect the outcome. Complete or partial synonyms include independent variable (IV), covariate, blocking factor, and predictor variable. 

    Clearly, classification of the role of a variable is dependent on the specific experiment, and variables that are outcomes in one experiment may be explanatory variables in another experiment. For example, the score on a test of working memory may be the outcome variable in a study of the effects of an herbal tea on memory, but it is a possible explanatory factor in a study of the effects of different mnemonic techniques on learning calculus.

“Most simple experiments have a single dependent or outcome variable plus one or more independent or explanatory variables.”

    In many studies, at least part of the interest is on how the effects of one explanatory variable on the outcome depends on the level of another explanatory variable. In statistics this phenomenon is called interaction. In some areas of science, the term moderator variable is used to describe the role of the secondary explanatory variable. 

    For example, in the effects of the herbal tea on memory, the effect may be stronger in young people than older people, so age would be considered a moderator of the effect of tea on memory. In more complex studies there may potentially be an intermediate variable in a causal chain of variables. If the chain is A ⇒ B ⇒ C, then interest may focus on whether or not it is true that A can cause its effects on C written only by changing B. 

    If that is true, then we define the role of B as a mediator of the effect of A on C. An example is the effect of herbal tea on learning calculus. If this effect exists but operates only through herbal tea improving working memory, which then allows better learning of calculus skills, then we would call working memory a mediator of the effect.

Classification by statistical type

    A second classification of variables is by their statistical type. It is critical to understand the type of a variable for three reasons. First, it lets you know what type of information is being collected; second it defines (restricts) what types of statistical models are appropriate; and third, via those statistical model restrictions, it helps you choose what analysis is appropriate for your data.

“Warning: SPSS uses “type” to refer to the storage mode (as in computer science) of a variable. In a somewhat non-standard way it uses “measure” for what we are calling statistical type here.”

Students often have difficulty knowing “which statistical test to use”. The answer to that question always starts with variable classification:

“Classification of variables by their roles and by their statistical types are the first two and the most important steps to choosing a correct analysis for an experiment.”

There are two main types of variables, each of which has two subtypes according to this classification system

·         Quantitative Variables

·         Discrete variables

·         Continuous variables

·         Categorical Variables

·         Nominal variables

·         Ordinal variables

    Both categorical and quantitative variables are often recorded as numbers, so this is not a reliable guide to the major distinction between categorical and quantitative variables. Quantitative variables are those for which the recorded numbers encode magnitude information based on a true quantitative scale. 

    The best way to check if a measure is quantitative is to use the subtraction test. If two experimental units (eg, two people) have different values for a particular measure, then you should subtract the two values, and ask yourself about the meaning of the difference. If the difference can be interpreted as a quantitative measure of difference between the subjects, and if the meaning of each quantitative difference is the same for any pair of values with the same difference (eg, 1 vs. 3 and 10 vs. 12), then this is a quantitative variable. 

    Otherwise, it is a categorical variable. Once you have determined that a variable is quantitative, it is often worthwhile to further classify it into discrete (also called counting) vs. continuous. Here the test is the midway test. If, for every pair of values of a quantitative variable the value midway between them is a meaningful value, then the variable is continuous, otherwise it is discrete. 

    Typically discrete variables can only take on whole numbers (but all whole numbered variables are not necessarily discrete). For example, age in years is continuous because midway between 21 and 22 is 21.5 which is a meaningful age, even if we operationalized age to be age at the last birthday or age at the nearest birthday.

“Measurements with meaningful magnitudes are called quantitative. They may be discrete (only whole number counts are valid) or continuous (fractions are at least theoretically meaningful).”

    Categorical variables simply place explanatory or outcome variable characteristics into (non-quantitative) categories. The different values taken on by a categorical variable are often called levels. If the levels simply have arbitrary names then the variable is nominal. 

    But if there are at least three levels, and if every reasonable person would place those levels in the same (or the exact reverse) order, then the variable is ordinal. The above examples of eye color and race are nominal categorical variables. Other nominal variables include car make or model, political party, gender, and personality type. 

    The above examples of exam grade, car type, and burn severity are ordinal categorical variables. Other examples of ordinal variables include liberal vs moderate vs conservative for voters or political parties; severe vs moderate vs mild vs no itching after application of a skin irritant; and disagree vs. neutral vs. agree on a policy question.

    It may help to understand ordinal variables better if you realize that most ordinal variables, at least theoretically, have an underlying quantitative variable. Then the ordinal variable is created (explicitly or implicitly) by choosing “cut-points” of the quantitative variable between which the ordinal categories are defined. 

    Also, in some sense, creation of ordinal variables is a kind of “super-rounding”, often with different spans of the underlying quantitative variable for the different categories. See for an example based on the old IQ categorizations. Note that the categories have different widths and are quite wide (more than one would typically create by just rounding).

IQ/Quantitative	0	20	50	70	90	110	140	200
IQ/Qualitative	idiot	Imbecile	moron	dull	Average	supreme	genius

    It is worth noting here that the best-known statistical tests for categorical outcomes do not take the ordering of ordinal variables into account, although there certainly are good tests that do so. On the other hand, when used as explanatory variables in most statistical tests, ordinal variables are usually either "demoted" to nominal or "promoted" to quantitative.

Tricky Variable Cases

    When categorizing variables, most cases are clear-cut, but some may not be. If the data are recorded directly as categories rather than numbers, then you only need to apply the “reasonable person's order” test to distinguish nominal from ordinal. 

    If the results are recorded as numbers, apply the subtraction test to distinguish quantitative from categorical. When trying to distinguish discrete quantitative from continuous quantitative variables, apply the midway test and ignore the degree of rounding.

    An additional characteristic that is worth paying attention to for quantitative variables is the range, ie, the minimum and maximum possible values. Variables that are limited to between 0 and 1 or 0% and 100% often need special consideration, as do variables that have other arbitrary limits.

    When a variable meets the definition of quantitative, but it is an explanatory variable for which only two or three levels are being used, it is usually better to treat this variable as categorical. 

    Finally we should note that there is an additional type of variable called an “order statistic” or “rank” which counts the placement of a variable in an ordered list of all observed values, and while strictly an ordinal categorical variable, is often treated differently in statistical procedures.