Presenting Data Gaps and Value Ranges

Summary

A single point value in an energy overview conceals whether it was measured, estimated or derived from a scenario. This interim report from the ongoing project P08 distinguishes three kinds of data, describes three display variants and uses a constructed example to show how data gaps should appear as gaps rather than as zeros. The task-based test plan is described; the empirical evaluation has not yet been carried out. There are therefore no final results. This report is a demonstration and not scientific evidence.

Research question

How can ranges of values and data gaps be made distinguishable in a compact overview? Two sub-questions follow: do readers recognise which kind of data they are looking at? And do they make well-founded decisions when values are missing or given only as a range? The first question can initially be examined on the display itself, the second only with the planned test.

Structure

Three kinds of data

• Measured value: a recorded value at a point in time. It, too, has some imprecision, which is not shown here for simplicity.
• Estimate: a value derived from other values for a time at which no measurement exists, given with a range.
• Scenario: a value under an explicitly stated assumption, usually for the future. It is neither measured nor a forecast but an if-then statement.

Three display variants

1. Single curve: one line connects all point values. Kind of data and range are not distinguished; missing values would be filled with zero or by interpolation. It serves as the baseline.
2. Differentiated time series: measured values appear as filled points, estimates as points with a vertical range, scenarios as a dashed line within a band. Missing hours stay empty and carry the label “no data”. The distinction is carried by shape and labelling, not by colour alone.
3. Table with a range column: each hour has value, range and kind of data in words in one row. Gaps are entered as “no data”. This variant is economical but shows no trend.

Design rules for all variants

Four provisional rules apply to the drafts and will be reviewed as the project proceeds. First, a gap appears as a gap and never as a zero. Second, every kind of data carries a label in words. Third, a range is an element of its own and is not folded into the point value. Fourth, a scenario is always named together with its assumption. The following table summarises how the three variants meet these rules.

Assumptions

• All figures are constructed and come from no measurement, grid or plant.
• The unit is abstract (units per hour, “U/h”); no energy carrier or plant is named.
• The ranges for the estimates and the scenario are set, not computed from data.
• The display variants exist as descriptions and drafts; an assessment by test participants is pending.
• The report is not aimed at consumers and gives no recommendation.

Illustrative examination

The table shows eight hours of a modelled day. It is a constructed example, not a measurement.

What the variants make of these rows can be shown with a simple summary figure. For the hours 06:00 to 12:00, five values exist (three measured, two estimated), and their sum is 288 U. If the gap in the single curve is treated as zero, the mean over seven hours is about 41.1 U/h. If the gaps are instead left out and explicitly named, the mean of the five available values is 57.6 U/h. The difference of about 16.5 U/h arises solely from how the gap is treated. Moreover, this mean mixes measured and estimated values; a careful report would show the two separately.

In the differentiated time series, 08:00 and 09:00 stay empty, and the scenario value at 13:00 is recognisable as an if-then statement. The table variant shows the same in words, without a trend. In this interim report, all three variants are only described, not tested.

Limitations

• The example has eight hours and one gap; real overviews are larger and more irregular.
• So far there has been only a structured walkthrough by the project team. It does not substitute for any statement about user performance.
• The ranges chosen are set; how ranges are correctly derived from models is a matter for the modelling projects.
• There is no consumer advice and no real user study.

Interim conclusion and next steps

The interim position is that the three kinds of data and the treatment of gaps can be described precisely and implemented in variants. Whether any of the variants actually supports readers better is open. The empirical evaluation has not yet been carried out; there is currently no result to report.

The planned task-based test plan provides for:

1. Task 1: name the kind of data of a marked point (measured value, estimate, scenario).
2. Task 2: state what is known between 08:00 and 09:00.
3. Task 3: derive a statement from the overview and justify it, even when values are missing.

The planned measures are correct distinction of the kinds of data, legibility, and the quality of the reasoned decisions. The structured walkthrough comes first. Follow-up questions arise with U12; the trial is set up in the Interaction Studio as a demonstration with task logs.

The tasks are to be presented with the variants in rotating order so that sequence does not distort the outcome. Sample, selection of participants and analysis have not yet been fixed. The protocol is to be described before data collection, not adjusted afterwards.