psychometrics

Measurement

When it’s hot out, how do you confirm your sweatiness? When you rent a car, how do you prove to the dealer your old enough? When your bartender announces the last call, how do you call bull? All of these encompass various ways we measure things in order to do other things. This process of measurement is so commonplace that we tend to overlook how much of our day is affected by measurement.

But what is measurement?

In as little terminology as possible, measurement is the process of assigning numbers to things, such that, we ideally classify them as accurate representations of characteristics of the thing we propose to measure.

Before I can discuss things like probability distributions, of statistical learning processes, it’s important to understand that the notion of measurement. Anything that I’d like to apply methods to need to be quantified such that I can use the pieces of information, in the form of numbers, to tell me something about regularities (or abnormalities) in our world.

Okay great, we need to meaningfully put numbers to things, but when we are using measurement in psychology that’s easier said than done. And in fact, I’d go as far as to say that in social science broadly are plagued by the issue of properly accounting for things in such a way that captures the essence of the concepts we are interested in studying.

A classic example is a depression.

Whether it be from knowing someone or having experienced it ourselves, we all have a sense of some of the characteristics that make up this idea of of depression. So how do we quanfiy it? Qulitativly, we could proabliy decsibe vidily some of the expereinces that come along with depression, but how do we put numbers to it? One apporach would be to count things we thing are realted. We could say eating less than average, spending time with a lower number of people, and a bunch more things.  But that’s just it. I just ended up using measurement to make some claim about how to best represent something via other sets of numbers.

I have been dancing around it, but the term I want to talk about is a construct. A construct is a complex thing. It’s a set of characteristics all thought to be somehow interrelated such that when they are brought together through some process we can meaningfully distill information about latent concepts we are interested in. In our example, depression would be the construct we are interested in measuring, and all of the other items listed would be ways we could indicate some change in signal for our construct of interest; from now on, let’s call these things indicators.

That’s it.

We measure things using numbers and the things we are most interested in measuring are sets of indicators that together make up constructs.

So how valid are our claims about our constructs? Are we really capturing  a representation of depression from something like levels of appetite? Furthermore, how reliable is our measurement? If a was able to measure depression a bunch of times for the same person, would I get the same score? Would it bounce around alot? 

In the future, I’ll talk more about these issues and introduce ways scientists have attempted to address them.

Leave a Reply

Your email address will not be published.