Measures are calculated values that use Data Analysis Expressions (DAX) formulas. They allow you to create insights into your data quickly.
All measurements involve comparison with some known quantity of the same kind. This comparison is called a measurement signal. This article is about the measurement signal, and how it relates to other measurements.
Definition
A measure is a raw data representation, for example a number or value. It is used alongside dimensions, which are categorical buckets that can be used to segment and filter data. Examples include sales, leads, distances, temperatures and weights.
The word is also used to describe an instrument for measuring or a unit of length, such as a ruler, albeit without the common meaning of “rule.” However, a ruler is usually called a rule and the special name straightedge is reserved for a calibrated device that can draw straight lines.
A measurement is considered artifact-free when it relies on a fixed value of a physical constant, such as the International System of Units (SI)’s seven base units of length, mass, candela, second, ampere and kelvin. This approach is the one most widely accepted.
Purpose
One of the central concerns of measurement theory is error. Whether physical or psychological, all measurements are subject to some kind of uncertainty or error. Even well-designed instruments cannot eliminate all error; that is why many physical measurements are accompanied by an indication of their likely degree of accuracy.
In psychological measurements, two salient characteristics of measurement tools and methods are reliability and validity. Reliability refers to the consistency with which a tool assigns scores to subjects, and validity refers to whether a measure is suitable for the purpose intended. Neither of these is easy to establish. For example, it is difficult to determine the valid score for a reading test unless it is sampled frequently enough to detect differences among individuals with differing amounts of knowledge about the content.
Variables
Essentially, a variable is anything that can be measured and can assume different values. Height, age, income, province or country of birth and grades obtained at school are all examples.
Variables are then classified into two larger categories – qualitative and quantitative. Quantitative variables are those that can be described in numbers – they are the kinds of things that you can crack out a ruler to measure.
Qualitative variables, on the other hand, cannot be quantified. They can, however, be ranked – hence the term ordinal. Ordinal variables can also be coded and given value labels – such as “Expgradate” for the expected date of college graduation. This allows you to perform calculations that have a greater level of sophistication than would be possible with just names.
Formats
A measure can have a format string that specifies its output. It can also have a numberFormatStyle that controls how numbers are output. For example, percent formats a numeric value with a percentage symbol and currency formats monetary values.
A formatting string generates human-readable text from a Measurement and it can be configured with its width, numberFormatStyle, usage, and locale parameters. The system caches unique configurations of the formatting string to improve performance.
You can change the format of a single measurement by using its context menu in either the Tables or Columns panel. However, formatting for multiple measures cannot be changed from the context menu. You can use dynamic format strings with local model measures, but not remote models or live connect reports.
Examples
Measures are used to perform simple summarizations and calculations on data. For example, if you select a visualization that displays numeric values such as SalesAmount, Power BI Desktop creates and calculates a measure to aggregate the data. You can also create and calculate your own measures through the Fields pane or with DAX.
Using a thermometer to read the temperature of a cup of coffee is an example of a measurement system. Thermometers are usually judged to be accurate within a certain range, or in this case within one degree of the actual temperature.
Another measure is the mode, which describes a frequency distribution by describing the most common value. It’s less precise than the median or mean, but it is easier to interpret.