Age as a variable: Continuous or categorical?

Sir,

In a strongly worded letter, Mondal[1] affirms that age is indeed a number that can be categorized in groups. I agree. In fact, I had stated at least 3 specific circumstances in which this may be desirable: For administrative purposes; when the data cannot be accurately recorded; and when the data are skewed.[2]

A classical example for the categorization of continuous data is for public health purposes, as when patients are classified into pediatric, adult, and geriatric groups for health-care resource allocation; and in an allied context, for everyday descriptive purposes, as when people are referred to as being in their early twenties or mid-forties. The classroom cartoon in the letter[1] reflects this latter situation. However, there is a world of difference between a classroom in which research is talked about in conversational terms and a research article in which precision is expected because inferential statistical tests will be applied to the data.

Mondal[1] suggests that age can be viewed as a discrete variable because it is commonly expressed as an integer in units of years with no decimal to indicate days and presumably, hours, minutes, and seconds. If so, would ratings of psychopathology on various clinical scales or ratings of cognitive performance on various neuropsychological tests be considered discrete outcomes because no decimal ratings are assigned? Certainly not. The difference between ordinal data and data that are measured using an interval or ratio scale does not depend on the actual values recorded; it depends on the potential of the scale to accommodate decimal values. Whereas there may not be a benefit in excessive precision in recording age in adults, age in children is often expressed in months rather than years, so that the desired precision can be obtained, and age in neonates is expressed in days for the same reason.

The suggestion that education be categorized is suitable for administrative or everyday descriptive contexts. It is incorrect when education is treated as an independent variable the effect of which on outcomes is intended to be studied using inferential statistical procedures. The reasons were stated in my earlier article.[2] Operationalizing education in years is particularly useful when subjects complete extra courses, training, or diplomas at a horizontal level without gaining an additional degree.

There is no challenge at all in presenting continuous and categorical data in the same table, along with the appropriate inferential statistical results. In fact, this is commonly done. Any journal of good standing will contain plenty of examples.

Concluding with a note of humor, as Mondal[1] displays in his letter: Would he state his age as, say, “21-30” years in a job application or would he state a precise number?

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.