Examples explained
Start with the first — it shows the key property of the method
Your data
What n and units mean
n is the number of discrete values the range of actual values (RAV) is split into; the vote chooses among them. Empty — as in the method: a partition norm h of the order of the narrowest interval, but at least 11 values. A larger n gives a smaller norm h and a smaller uncertainty component from the partition. If the winners are not adjacent, the method increases n itself.
Units are used only for labels on charts and in the report; they do not affect the calculation.
Load a CSV / TXT file and templates
A file holds the same lines as the field above: one line per interval, in the order of the line format. The first line may be a header — it is skipped.
| Line | Separator | Where from |
|---|---|---|
Lab-1,10.012,0.010 | comma, decimal point | English Excel, Python, R |
Lab-1;10,012;0,010 | semicolon, decimal comma | European Excel |
Lab-1 10.012 0.010 | space or tab | TXT, copied columns |
Templates — open in Excel, replace the rows with your data, save and load back:
The file is read in your browser and sent nowhere. Save XLSX files as CSV.
Preferential median
Grey intervals do not contain the PM. “± u” is the standard uncertainty of the PM: the partition component 0.41h and the sampling component (details in the step-by-step view).
Experiment with these data
Move an interval, make it narrower or wider — the PM, the mean and the weighted mean update at once. The original data stay unchanged until you press the button below.
What it means
Comparison with other estimators
The same data, ten ways to get one value. “Difference from PM” is how far an estimate lies from the preferential median. When each estimator is stronger — see the configurator.
Coverage
How many intervals cover each discrete value of the RAV; Borda winners are highlighted.
Step by step
Profile of interval-induced rankings
Each row is the “ballot” of one interval: blue cells are RAV values inside the interval (preferred), grey ones are outside. At the bottom is the coverage; orange bars are the winners.
Table of intervals
“(x − PM)/u” is how many of its own uncertainties a result lies from the PM; above 2 in absolute value — the result clearly disagrees with the consensus.
Share and ask
The buttons open your e-mail program with a ready message — the site neither sends nor sees it. The data are stored in the link itself; before sharing, make sure they contain nothing you may not disclose.
The script repeats this calculation with your data; it needs the file from the
Code page next to it (ifpa.py or ifpa.R). The expected result is written at the top of the script.
How to cite. If the result goes into your work, please cite the paper describing the method: Muravyov S.V., Khudonogova L.I., Emelyanova E.Yu. Interval data fusion with preference aggregation. Measurement, 2018, 116, 621–630. doi:10.1016/j.measurement.2017.08.045 BibTeX and other formats →