Regression analysis

Regression analysis is a static tool used to identify trends or outliers graphically. Two measured channels that are directly related to one another are required, for example electricity (kWh) and production quantity (units). Based on their temporal assignment, a point is generated in the coordinate system for each value pair. All data points (pairs of values) for a given time range can then be plotted on a chart.

For further analysis, a line can be generated using a selected statistical function. Seven different statistical methods are currently available. Using the line, corresponding values on the second axis can be derived for theoretical values on the first axis. Optionally, an R² value indicating the goodness of fit of the selected statistical function can be displayed.

Any measured channels (metering point / OBIS) can be related to one another.
Seven different statistical functions to choose from (linear, quadratic, polynomial, exponential, logarithmic, power, Loess)
The colour and line thickness of the regression line can be selected.
Point size can be set freely.
The R² goodness-of-fit value can be displayed.
All standard chart functions (selecting a time range, PDF printing, use as a favourite, and much more) are available.

Chart properties

Axis font	Selection – Select the colour, font and size for the axis labels.
Legend	Specifies which columns are displayed in the chart legend.

Toolbar

Chart
	Refresh	Reloads the data from the database and updates the chart. Please note: any unsaved changes will be overwritten by the most recent version stored in the database.
	CSV download	Generates a CSV data file and enables the browser’s download/save function.
	Properties	Opens a dialog box containing the chart properties. Here, you can, among other things, show or hide the columns in the legend.
Display (not available for all chart types)
	Time chart view	Values for the current time range are displayed in a level (x)–time (y) view. Depending on the measured variable, the level can be displayed as a line, a bar or point.
	Value table view	The values for the current time range are displayed as a table.
	Frequency distribution view	Values in the current time range are sorted by value rather than chronologically by time.
Time range
	Day view	The time range is scaled to one day.
	Week view	The time range is scaled to one week.
	Month view	The time range is scaled to one month.
	Quarter view	The time range is scaled to a quarter.
	Year view	The time range is scaled to one year.
	Display entire time range	The entire time range available in the database is loaded and the view is scaled accordingly.
	Custom time range	The display range can be customised by specifying a start and end date.
Navigation
	Go to the first record	The first available data is loaded and displayed within the currently selected time range.
	Go to the previous range	Displays the previous range.
	Go to the next range	Displays the next time range.
	Go to the last record	The most recent available data are loaded and displayed within the currently selected time range.
Show legend
Allows you to show or hide the legend.

Channel properties

Name/Designation/Title	A meaningful text entry. The designation is used for identification and search purposes and can be changed at any time.
Enabled	Yes / No – Enables or disables the object.
Colour	Selection – Allows you to customise the colour for the channel.
Rank	Number – Allows you to set the order of the channels. The order can be crucial when elements overlap. This property can also be changed by dragging the channel with the mouse. The first channel is ranked 0, and so on.
Point size	Number 1–100 – Allows you to set the size of each data point individually.
X-axis	Display – Shows the currently defined measured variable with the X-axis assignment.
Y-axis	Display – Shows the currently defined measured variable with the Y-axis assignment.
Swap axes	Action – Swaps the current X and Y axis assignments.
Show tooltip	Yes/No – A tooltip showing the value, time and status can be displayed for each data point when the mouse hovers over it. When displaying larger time ranges with a large number of data points, it is advisable not to enable this option.
Show regression	Yes/No – The regression line and result derived from the data points, time range and regression function can optionally be displayed on the chart.
Colour	Font colour selection – The font and font size are set by the system.
Line thickness	Number 1 – 20 – Defines the thickness of the statistical regression line.
Regression function	Selection – Linear – A mathematical model that describes the relationship between a dependent variable (e.g. consumption) and an independent variable (e.g. production quantity) in the form of a straight line. Recommended for static relationships. Quadratic – An extension of linear regression analysis in which the relationship between the dependent variable and the independent variable is not linear but curved in the shape of a parabola. Recommended for cost trends (costs falling initially and then rising later). Polynomial – A generalisation of linear and quadratic regression. It describes the relationship between a dependent variable and an independent variable using a higher-order polynomial. Recommended for non-linear trends in time series, such as seasonal fluctuations. Exponential – A regression model used when the dependent variable increases or decreases exponentially with the independent variable. Recommended when small changes have proportionally larger effects. Logarithmic – A regression model used when the dependent variable y is related to the independent variable x in such a way that y changes rapidly for small values of x, but more slowly for larger values of x. Recommended for processes with diminishing growth. Power – A regression model used when the relationship between two variables can be described by a function of the form y = a⋅x^b. y: Dependent variable x: Independent variable (must be positive) a: Scaling factor (initial value) b: Exponent (indicates how strongly y responds to changes in x) Loess – A non-parametric regression method used to fit smooth curves to scatter plots, particularly when no clear mathematical relationship between the variables is known. Example: measurement data on temperature trends throughout the day where the values fluctuate significantly. A LOESS curve can help to illustrate the general trend (e.g. rise in the morning, plateau at midday, fall in the evening) – without the need to assume a specific equation.
Show R²	Selection Yes / No – The R² value (coefficient of determination) is a key measure in regression analysis. It indicates how well a regression model explains the variation in the data – in other words, how well the regression function fits the observed data. R²=1: Perfect fit – the model explains 100 % of the variation in the data. R²=0: The model explains nothing at all – it is no better than the average. 0 < R² < 1: Partial explanation – the closer to 1, the better. R² < 0 (rare, but possible with non-linear models): The model performs worse than a horizontal line (e.g. the average).
Show formula	Selection Yes / No – The selected function is displayed as a formula in the chart.
Font size	Number – Defines the font size for R² and the formula.
Position	Selection – Defines the position for R² and the formula Top right Top left Bottom right Bottom left