Wintrack

Transform statement

 Transform
   Mode=RunAve|RunDiff|RunSum|BackSum|Abs
     |Diff|Step|DivTim|DivPath|MultTim|MultPath|Equalize
     |Zero|MissVal|Fill|Frame
     |ClusterSum|ClusterAve|ClusterMax|ClusterMin
     |ClusterDir|ClusterDivers|ClusterSpread|ClusterNeigh|MinDist
   [Frame=frame_value]

Type conventions and statement syntax are explained in the Glossary.

Apply the transformation indicated by the Mode parameter to the cell values
in all columns of the active data matrix.

Parameters...

Mode
Radio button parameter indicating the transformation mode. There is no default, specify one of the following options:

RunAve = running average using the time window specified by Frame. If there are empty cells within the averaging window, the transformation result is the average of the remaining non-empty cells. If all cells within the averaging window are empty, the result is an empty cell.

RunDiff = replace the value of each cell by the difference of that value to the value in the preceding cell. The first cell will become empty, as well as all cells preceded by an empty cell. Empty cells remain empty. The function performs a circular transformation of the difference values if parameter Frame>0: e.g. when treating differences between direction values with Frame=360 a difference of 270 is replaced by -90.

RunSum = replace the value of each cell by the sum of the values of all non-empty cells from the beginning up to that cell. Empty cells remain empty.

BackSum = replace the value of each cell by the sum of the values of all non-empty cells from the end back to that cell. Empty cells remain empty.

Abs = replace all negative cell values by corresponding positive values. Empty cells remain empty.

Diff = cells are grouped into contiguous clusters with same value: cluster boundaries are either defined by value changes or interveening empty cells. The first cell of each cluster is set to the difference between its value and the value of the closest upsteram cluster. All other cells are emptied. This mode is intended for calculation of relative field indices in data matrices representing field entries (Matrix statement with Data=Field). The function performs a circular transformation of the difference values if parameter Frame>0: e.g. when treating differences between direction values with Frame=360 a difference of 270 is replaced by -90.

Step = as Diff, but differences are calculated downstream.

DivTim = divide the value of each cell by the time stamp difference between the corresponding and the nearest upstream data point. Empty cells remain empty.

DivPath = divide the value of each cell by the length of the path segment between the corresponding and the nearest upstream data point. Empty cells remain empty.

MultTim = multiply the value of each cell by the time stamp difference between the current and nearest upstream data point. Empty cells remain empty.

MultPath = multiply the value of each cell by the length of the path segment between the current and nearest upstream data point. Empty cells remain empty.

Equalize = divide the value of each cell by the sum of all cells in the column and multiply the result by 100. Empty cells remain empty.

Modes above are not affected by the selection state of the cells in the matrix. Below are a set of modes that transform the values of or generate new values for selected cells. Deselected cells remain unchanged:

Zero = set all deselected non-empty cells to zero. Empty cells remain empty.

MissVal = empty all deselected cells, that is assign them a missing value code.

Fill = set all deselected cells to value of next upstream selected cell.

Frame = replace the value of all selected cells by the one supplied with the Frame parameter.

ClusterSum = look for clusters of selected cells. For each cluster found, calculate the sum of all non-empty cells in the cluster and write that sum to all cells in the cluster, including to empty cells. If all cells of a cluster are empty they remain empty.

ClusterAve = look for clusters of selected cells. For each cluster found, calculate the average of all non-empty cells in the cluster and write that average to all cells in the cluster, including to empty cells. If all cells of a cluster are empty they remain empty.

ClusterMax = look for clusters of selected cells. Within each cluster found, replace all cell values (including empty cells) by the largest value in the cluster. If all cells of a cluster are empty they remain empty.

ClusterMin = look for clusters of selected cells. Within each cluster found, replace all cell values (including empty cells) by the smallest value in the cluster. If all cells of a cluster are empty they remain empty.

ClusterDir = look for clusters of selected cells. Within each cluster found, replace all cell values (including empty cells) with the direction angle of the vector connecting the first point in the cluster to the last point in the cluster. If all cells of a cluster are empty they remain empty.

ClusterDivers = look for clusters of selected cells. Within each cluster, replace all non empty cells by the diversity of the corresponding cluster points, that is the average distance between any two points, weighted by the distribution of the cell values.

ClusterSpread = look for clusters of selected cells. For each cluster determine the center of gravity of the corresponding points. Then replace all cell values (including empty cells) in the cluster by the distance of the corresponding point to the cluster's center of gravity.

ClusterNeigh = look for clusters of selected cells and replace all cell values (including empty cells) by the sum of first values of the corresponding point's neighboring clusters. Neighboring clusters are those that contain at least one point whose distance to the examined point is smaller than the value specified by the Frame parameter.

MinDist = Replace all cell values by the distance of the corresponding point to the nearest selected point. By definition, all selected cells obtain value 0.

Frame
Number parameter needed if RunAve, Frame, or ClusterNeigh are used for Mode. For RunAve, it specifies the width of the averaging window in seconds. For Frame it specifies the value to which selected cells should be set. For ClusterNeigh it specifies the maximally allowed distance for neighbors. If RunDiff or Diff is specified for Mode, this parameter is optional: values >0 trigger a circular transformation of the difference values, e.g. when treating differences between direction values with frame=360 a difference of 270 is replaced by -90.

Note that...

Transformations never modify the active selection mask. Except for Mode=Zero, Mode=MissVal, Mode=Fill, Mode=ClusterSum, and Mode=ClusterAve, they are not affected by the active selection mask. They are applied to all cells of the active data matrix, no matter whether they are currently selected or deselected. You may, but do not have to, apply transformations to the active data matrix before using it to create variables.

Additional information...

Managing custom variable definition files provides general information on the management of custom variable definition files.
The Flowchart of advanced data analysis gives a synopsis of data and variable types.
There is a topic describing how to build a custom variable definition file.
There is a list of all available variable definition statements.
Using the Memory statement you can build collections of data matrices and selection masks for later reuse in other variables.
See Analyze Case Document command for basic information on that command.
Inside case documents provides advanced information on case documents.