Analysis
Frequency Analysis
How the histogram is binned, what the count, relative-frequency and density views show, and how to read the frequency table and the frequency-by-period heatmap.
Histogram
The histogram groups the values of the current selection into classes (bins) of equal width. Each bin includes its lower edge and excludes its upper edge: a value v belongs to the bin with x₀ ≤ v < x₁. The edges are chosen so that the last bin ends above the maximum value, so every value is counted exactly once.
Views
| View | Bar height | Use it to |
|---|---|---|
| Count | Number of values in the bin | See absolute volumes |
| Relative frequency | count / n | Compare selections of different size; bars sum to 100% |
| Density | (count / n) / bin width | Compare shapes across different bin widths; the area of all bars equals 1 |
Mean and median markers
Vertical markers show the mean and the median of the selection. When the mean lies to the right of the median, the distribution usually has a longer right tail (positive skewness); the Support Response Times dataset is an example. The tooltip of a bin shows its class boundaries, its count and its share of n.
Bin rules
For continuous data, the bin width h is first calculated with the selected rule, using the n values of the selection, their range R = max − min, the interquartile range IQR and the sample standard deviation s.
Readable edges
The raw width is then rounded up to the nearest value of the form 1, 2, 2.5 or 5 × 10ᵏ (for example 0.13 becomes 0.2, 7 becomes 10 and 23 becomes 25). The first edge is the largest multiple of the width that is not greater than the minimum value, and edges continue in steps of the width until one exceeds the maximum.
Fixed bin count is a target
Because the width is rounded to a readable value and the edges are aligned to multiples of it, a fixed bin count of k can produce a slightly different number of bins. The number of bins is capped at roughly 400.
Discrete data
For datasets with value kind discrete (such as Items per Order), the histogram uses one bin per integer, centred on the integer: [k − 0.5, k + 0.5). The bin rule selection does not apply. If the integer values span more than 60 distinct integers, the continuous rules are used instead to keep the chart readable.
Frequency table
The frequency table lists the same classes as the histogram, from lowest to highest, including classes with a count of 0.
- Class
- “lower – upper” with the dataset's recording precision; for discrete data, the integer value (for example “3”).
- Absolute frequency
- fᵢ = number of values in class i.
- Relative frequency
- fᵢ / n.
- Cumulative frequency
- Fᵢ = f₁ + … + fᵢ, the number of values in class i or below.
- Cumulative relative frequency
- Fᵢ / n. The last row is always 100%.
Read the cumulative relative frequency to answer questions such as “what share of observations had demand below 400 MWh?”: find the class whose upper edge is 400 and read its cumulative relative frequency. The answer is exact only at class edges. The table can be exported as CSV from Reports.
Frequency by year or quarter
The heatmap shows how the distribution changes over time. Rows are periods (years such as 2024, or quarters such as 2024 Q1); columns are the same classes as in the histogram, computed once from the whole selection so that every period uses identical bins.
Because each row is normalised by its own number of observations, periods with different numbers of observations remain comparable; a partial year at the edge of the date range is not penalised for having fewer days. A shift of the more intense cells to the right from one period to the next means that higher classes became more common.
Small periods
A period with few observations can show large shares in single classes. Check the number of observations per period before interpreting differences, and prefer years over quarters for short date ranges.
Empty selections
When the selection contains no observations, the histogram, frequency table and heatmap have nothing to show. When all values are identical, the histogram has a single bin. Adjust the filters as described in Data Explorer. For the distribution shape per category, continue with Distributions.