Regression analysis on data in a Conversion Table

To perform regression analysis on two variables a conversion table must exist (see Create a Conversion Table).

1. Click Regressions on the DataSight Ribbon. The Regressions Window opens.
2. On the Regressions window, view the list of conversion tables by clicking Conversion Table on the Menu Panel. The Conversion Table window opens.
3. To load a conversion table, click on its row and then click OK. The table will be displayed in the Conversion Table panel as well as graphically in the Regression Chart.
4. Click Gaugings in the Regression Group. Click OK to load all the Gaugings or select the desired Gaugings by checking the box(es) against each row, then click OK. To view the coordinates of a gauging point on the chart, hover your mouse over it and hold to display its label.
5. To develop a best "fit" Conversion Table, follow the iterative process below:

Develop a Best fit Conversion Table

1. Click Gaugings in the Regression group.
2. Load all the Gaugings or select the desired Gaugings by checking the box(es) against each row.
3. Then click OK.
4. To view the coordinates of a gauging point on the chart, hover your mouse over it and hold to display its label.
5. Right click within the Chart to show the Regressions Chart context menu.  All series are shown on the chart by default.  Click on each series individually to toggle their display setting on or off.  You can also Transpose the axes, Copy the chart to clipboard, or Print the chart directly.                                          
6. (Optional) Click on the Confidence Chart This shows the conversion table as the zero point and the Gaugings as a percentage off the curve.
   • The quality of the Gaugings is also represented by scaled error bars - the larger the error bars, the lower the quality of the gauging measurement.
   • The Confidence chart is interactive. When you hover the mouse over a gauging, details regarding the gauging are listed.
   • You may also add confidence values from other Conversion Tables by clicking on Conversion Table to load them and electing to switch to Confidence Chart. Duplicated values are ignored in the output.
7. (Optional) To exit the Confidence Chart, click Regression Chart.
8. (Optional) Click Shift to perform a shift analysis (See Shift Analysis).
9. (Optional) Click Transpose to switch the axes of your curve. You can also right-click on the chart and select Transpose.
10. (Optional) You can choose to copy values from the original table in the left panel into your output. This is helpful when only a subset of the Conversion Table rows needs to be manipulated. To do this, right-click on a row you wish to copy to the output and select Copy Values to Output. To select multiple rows or a range of rows, hold down CTRL or SHIFT respectively when clicking on the rows. Then copy values with the right-click menu.
    • It is possible to ignore the copied rows from affecting your curve (as they are already added to the output), by right-clicking on the row(s) and electing to Unselect. To include an unselected row back into your curve, right-click on the row and click Select.
11. (Optional) Click Criteria to apply start and end values on your Conversion Table x-axis and output increments for the regression analysis. You can elect to:
    • Automatically set start and end values: the regression tool will use the lowest and highest values in the data as the start and end values. For a transposed curve, the start and end values apply to the original x-axis, not the x-axis post-transpose.
    • Check Use these values and enter values manually.
    • Enter a number of Increments at each step of the regression or enter a total Number of increments for the regression analysis.
    • Click OK to load your preferences.
12. Click Analyse on the Regressions Ribbon Menu to model your curve against the Regression Models in DataSight. A list of Regression Models will open in a new window. Click Analyse to confirm you wish to carry out the analysis, otherwise click Cancel.
13. Select models on the right by clicking on them to find the best fit for your curve. The regression analysis will be completed using all the regression models according to a goodness of fit and the results are displayed to the right of the Regressions window. Selecting a model from the list will display the model graphically and present the R2 value.
14. Review the results, then select or unselect points on the original table and/or gaugings to achieve a better "fit".

See also:

• Save a Conversion Table
• Regression Settings
• Regression Models
• Regressions Layout
• Conversion Tables
• Shift Analysis

Save a Conversion Table

You can model your data using Regression Analysis and then save the results as a Conversion Table.

Save a Conversion Table

1. Perform a Regression Analysis.
2. Select the Results tab to display the finalised results of the selected model.
3. Highlight the value(s) from the Results that you wish to move across to the Output Table, right-click and select Copy Values to Output. Data in the Output tab is displayed in the Chart area in blue.

                To select multiple values, hold down CTRL or SHIFT and click on the rows.

4. The Output table will be displayed in the charting area and can be saved (as a DataSight Conversion Table) by clicking Save on the Quick Access Toolbar and entering the Conversion Table Name, the X and Y axis Labels, and the Period of Applicability (this is simply a set of reference dates for the user). To prevent overwriting of an existing table, the system date is automatically appended to the Conversion Table name upon saving. You may edit this as you wish. You can also assign the conversion table to a Site (Level 3).
5. When the relationship between your variables changes, you can elect to use differing regression models to build consecutive parts of the Conversion Table. Simply copy the selected rows of data from the Analyse Results Tab into the Output Table before saving.
6. You can add Comments about the new Conversion Table as required.
7. Click Save to save the Conversion Table. The Conversion Table will then be available for interpolation and use with other datasets.
8. You can construct the one conversion table from consecutive data ranges that vary in their modelled relationships. In this instance, you need to reiterate the process of including/excluding data points from the analysis and once each section is modelled, copying that regression data to the Output table as explained in step 2.

Data Conversion

Applicable data can be converted immediately after saving the Conversion Table.

1. When you Save a Conversion Table, check the Apply to existing data box prior to saving the Conversion Table.
2. The Data Conversion window opens. Select an Input Variable. The values for this variable are converted according to the saved Conversion Table.
3. (Optional) Set an input Flag.
4. Click Load Data to view conversion results on a chart on the right.
5. Select an Output Variable from the drop down list to save converted results to. If the variable does not already exist on the level, it will be created there by appending to the database.
6. (Optional) Set an output The default setting will Keep flag as it is from the input variable.
7. Select the number of Decimal Points that you wish to use to format your Output Variable values. The default setting is 1 decimal place.
8. Click Save to apply conversion to the data. The progress bar runs until the process has completed.

Data can also be converted during Calculations using saved Conversion Tables (See Use a Conversion Table for more details).