- Feature Engineering Made Easy
- Sinan Ozdemir Divya Susarla
- 251字
- 2021-06-25 22:45:55
The interval level
We are starting to cook with gas now. At the nominal and ordinal level, we were working with data that was qualitative in nature. There was data that did not describe a true quantity. At the interval level, we move away from this notion and move into quantitative data. At the interval data level, we are working with numerical data that not only has ordering like at the ordinal level, but also has meaningful differences between values. This means that at the interval level, not only may we order and compare values, we may also add and subtract values.
Example:
A classic example of data at the interval level is temperature. If it is 90 degrees in Texas, and 40 degrees in Alaska, then we may calculate a 90-40 = 50 degrees difference between the locations. This may seem like a very simple example, but thinking back on the last two levels, we have never had this amount of control over our data before.
Non-example:
A classic non-example of data that is not at the interval level are Likert scales. We have identified Likert at the ordinal levels for their ability to be ordered, but it is important to notice that subtractions do not have a true consistent meaning. If we subtract a 5-3 on a Likert scale, the resulting 2 doesn't actually mean the number 2, nor does it represent the category 2. Thus, subtraction in a Likert scale is difficult.