Tip: While the audio is playing, you should take notes. Try to identify main points if possible. After listening to the audio clip, you will be given five to six multiple choice questions. You will not be given a transcript or be allowed to listen to the recording again.   Narrator: Listen to the following portion of a lecture and discussion from a statistics class. Male Professor: So, recall that we’ve been looking at the hierarchy of the four scales of measurement we use in research, which will be important for you to consider as you design your own experiments and evaluate research.... Show more

Tip: While the audio is playing, you should take notes.

Try to identify main points if possible. After listening to the audio clip, you will be given five to six multiple choice questions. You will not be given a transcript or be allowed to listen to the recording again.

 

Narrator: Listen to the following portion of a lecture and discussion from a statistics class.
Male Professor: So, recall that we’ve been looking at the hierarchy of the four scales of measurement we use in research, which will be important for you to consider as you design your own experiments and evaluate research. As a quick review, we have nominal, ordinal, interval, and ratio. Nominal scales are the lowest level of measurement. We also refer to them as classificatory scales, wherein objects or people are assigned to categories according to some criterion. Ordinal scale measurements require categories to be rank ordered on the basis of an operationally-defined characteristic or property. For example, customer satisfaction ranked 1-5. Interval scales possess the rank order characteristics of an ordinal scale but there are known equal distances between consecutive units of measurement. This allows relative differences in equivalences within a scale to be determined. Ratio scales achieve the greatest measurement specificity. A ratio scale is an interval scale with an absolute zero point that has empirical, rather than an arbitrary, meaning.
Today I want us to consider the relevance and importance of assessing accuracy and precision within each of these scales. Before we do this though, who is willing to remind us about what accuracy and reliability mean? Yes, Janet.
Female student: Well, uh…accuracy refers to the closeness of a measured value to a standard or known value. For example, hmmm… if you obtain a weight measurement of 3.8 kg for a given substance, but the actual weight is 10 kg, then your measurement is not accurate. Precision, on the other hand, refers to like the closeness of two or more measurements of the same item to each other. Using the same example, if you weigh a given object five times, and get 3.6 kg each time, then your measurement is very precise.
Male professor: Very well said, Janet and those examples were perfect. And remember, precision is independent of accuracy. You can be very precise but inaccurate and you can be accurate but imprecise. So now, let’s try to layer this thinking onto our different measurement scales and evaluate them through the lens of their potential accuracy and precision. I’ll start with nominal scales and then see if anyone wants to take a stab at any of the others. Ok? Because nominal scales are more like categories, it is hard to have much precision. For example, if the categories are various colors, our values could be blue, red, and green. But, the categories will have many shades of each color that all have a common general color but many variations of the hue. Sky blue and navy look very different, right? Accuracy would also be challenging. Continuing with our color example, the delineations between the categories are not very specific. There can be shades of blue that also appear green such as teal, so which category would they go in? One investigator could select green while another chooses blue.
So, let’s consider ordinal scales. Remember, they are represented by rank orders divided by intervals that are not always consistent or known. Two subjects assigned in the same rank may in fact be of completely different values. The ordinal scale is not sensitive enough to determine the differences between ranks and is only able to indicate a relative position of certain distribution rather than its true value or quantity. Manual muscle test is an ordinal scale test. Assigning a grade to an individual may be very accurate and precise based on the scale. However, this scale is pretty much useless when comparing two or more individuals since the difference between two scores is hard to be defined. Who is willing to stick their neck out and try interval scales?
Male student: Interval scales are essentially an intermediary between the lack of specificity of nominal and ordinal but a little less defined than ratio with no absolute zero, so I think we cannot be quite as precise as ratio but more than the lower levels of measurement. My hunch is that the artificial zero point that these scales possess decrease its accuracy because the zero points are just arbitrarily chosen.
Male professor: You hit the nail on the head, Donovan! Who wants to try ratio?
Female student: Well, I think because there are essentially an infinite number of values along the continuum in the ratio scale, it is easier to be more precise because the difference from one value to the next is very, very small. I’m not sure about accuracy.
Male professor: Great start. You’ve hit on an important concept. In general, scales of higher forms of measurement are more precise. Accuracy is also more important in scales like interval and ratio that are higher levels of measurement and may have more specificity to their assigned values. In a sense, the target on the bullseye is smaller, so it seems logical that it would be easier to miss it and be inaccurate more frequently than in large sweeping general categories like nominal and ordinal scales. For instance, we know gender is nominal. It’s generally easy to identify someone as male or female so you are less likely to make a mistake. Salary is a ratio measurement and since there are essentially an infinite number of salaries, it would be easier to misreport it and have an accuracy error. Perhaps you inadvertently flip the numbers around, or mishear a fifty as sixty.

Show less