Filter Case Document command

How to filter a case document...

1. If you wish to add a report of the filter operation to an existing scrollsheet, make sure it is open and placed on top of any other open scrollsheet. If you wish to write a filter report to a new scrollsheet, Wintrack will create it automatically.
2. Open the case document you wish to operate on and make sure it is the active document. Use the mouse to select the trials you want to filter.
3. Select Filter Case from the Analysis Menu.
4. A dialog box appears. Press the Help button if you need instructions on how to fill in the dialog box fields. Fill in the fields that are not grayed or accept defaults and press Ok to continue. Or press Cancel to abort.
5. Wintrack will filter the active case document. It will also write a report of the filter operation to the topmost open scrollsheet or create a new scrollsheet, depending on the selections you made in the dialog.
6. You may repeat steps 2.-5. as many times as you wish to accumulate filter reports from several cases in the same scrollsheet.

Filtering cases is usually not necessary. However, path data may contain artifacts due to electrostatic discharges, weak contrast of the tracked object, or uneven background. There may as well be gaps in the data due to weak contrast or diving of an animal in a water maze. Since gaps and artifacts in the data may affect analysis results, Wintrack provides this command which automatically detects and eliminates minor artifacts. Some fine tuning of the filtering thresholds and parameters may be necessary for good results. Be cautious! Using the command with inappropriate parameters may affect analysis results more badly than keeping a few artifacts in the data.

How Wintrack filters data...

Modeling a moving animal

Internally, Wintrack implements a mathematical model of a moving animal. This model is used for both path reconstruction and artifact detection. When checking for artifacts, Wintrack predicts the position of data points based on the time stamps and coordinates of down- and upstream points. The spatial deviation of the real from the predicted position is used as the criterion for artifact recognition.

Point deviations

In a first pass, the filter routine assesses the deviation of the real from the predicted position of every single data point. Whenever this deviation exceeds the limit specified in the field Point deviation: inactivate if in the Filter Case dialog, the point is discarded.

In a second pass, the position of each discarded data point is tested again. But this time, the prediction of its position is based only on data points which survived the first pass of filtering. If the deviation of the real from the predicted position of a previously discarded data point now falls below the threshold specified in the field Point deviation: restore if of the Filter Case dialog, the point is restored.

Segment shifts

If a series of consecutive data points forming an path segment is shifted, all data points in the shifted segment will be discarded, though the spatial relations within the segment itself may appear normal. This will happen provided that the shift exceeds the threshold specified in the field Segment shift: inactivate if of the Filter Case dialog and provided that the segment fits within the time span specified in the field Segment shift: ignore if.

Data reconstruction

Since Wintrack data analysis does not rely on a constant sampling interval, deleting data points does not directly affect data integrity. You may, however, specify a maximal time interval between subsequent data points using the field Point resampling: add if > [s] of the Filter Case dialog. If any gap in the data exceeds this limit, Wintrack will provide a reconstruction of the path across the gap. Scaled XY coordinates and or supplemental data channels will be reconstructed as well. The algorithm uses balanced inter/extrapolation to provide smooth reconstruction of xy data. By default, supplemental data channels are reconstructed using an algorithm that strongly emphasizes interpolation over extrapolation. You may, however, specify a range of channels that are to be reconstructed using balanced inter/extrapolation like XY coordinates. If the duration of the gap which is to be reconstructed exceeds the value specified in the field Segment shift: Ignore if > [s], it is filled using pure linear interpolation.

Redundancy filters

Having many data points that lie close together takes up a lot of storage space but provides little information. You can use distance or time interval between points as criteria for removal of unneeded points. In the field Point resampling: del if < [s] of the Filter Case dialog, you can specify a minimal point to point time interval. This is checked at the same time as the maximal interval specified for data reconstruction: a point is removed if the interval to the preceding point is too small, new ones are inserted if it is too large. Using field Point resampling: del if < [m], you can specify distance-based redundancy thresholds for point to point distance and/or deviation from a straight line. Wintrack will remove all points that lie closer than this distance to each other or to a straight line. This is done after point reconstruction has been completed.

Counting failures

After all deviated data points and shifted segments have been discarded and all gaps have been filled according to the specified parameters, Wintrack tests remaining and reconstructed data points a third time, again comparing predicted and real coordinates of each data point. All points whose deviation still exceeds the limit specified in the field Point deviation: failure of the Filter Case dialog, are counted as filter failures. You may have Wintrack automatically discard entire trials if the number of filter failures reaches the threshold specified in the field Trial Rejection: Threshold of the Filter Case dialog.

Manually marked points

There are keyboard shortcuts to manually mark data points while stepping through a zoomed case document before running the Filter Case Document command. Whenever the automatic filtering engine is unable to recognize artifacts correctly, you can use this feature to help Wintrack to decide which points to reject and which ones to keep. Three types of marks can be set. Points marked for rejection will always be rejected, while those marked as protected will never be rejected. If the document has an event channel, points marked as missing data will have the "Alt" event key set pressed and will be moved to the position of the last valid point, else they will be treated as marked for rejection. Manual filtering marks are cleared as soon as filtering is completed.

Related commands...

Use the Open Case Document command to open or import case files. Use the Open Scrollsheet command to load an existing scrollsheet. to which you may want to add further filter reports.

Additional information...

• See Filter Case Document dialog for details on the associated dialog box.
• The related macro statement is explained in Filter Case statement.
• There are toolbar shortcuts to filter case documents.
• Inside case documents provides advanced information on case documents.
• Inside scrollsheets provides advanced information on scrollsheets.
• Consult Creating and using macros to learn how to use and create macros.
• Keyboard Shortcuts gives an overview on keyboard shortcuts.
• Using the Wintrack toolbar gives an overview of toolbar shortcuts.