Dimensions

A Dimension is a set of segments (answer options). For example, “Gender” could be a dimension in a matrix, and within that dimension could be the segments “Male” and “Female”.

Go to the Sample Management > Current Matrix window, and open the Dimensions tab. This tab lists the dimensions registered for the currently selected matrix - .

Figure 1 - Example of the Dimensions tab list

In the example, the dimension is named Gender/Carbrands/Degree, and this is a "combined" dimension; one dimension that comprises two or more individual dimensions. The toolbar buttons are:

• Delete – deletes selected Dimensions.
• Move Up/Down – moves selected Dimensions up or down in the list.
• Combine Dimensions - click to combine the selected dimensions.
• New Dimension– creates a new dimension for the current matrix.
• Refresh– refreshes the display. This may be required in the event changes have been made to the list that are not displayed immediately.

Click on a Dimension Name link to open the Edit Dimension page for that dimension. The figure below - shows the details of the Degree dimension from the combined dimension shown above.

Figure 2 - Example of the Edit Dimension page for a dimension

The toolbar buttons and data fields are:

• Save – saves changes for the tab. The button flashes while Forsta Plus has unsaved changes.
• Back to Dimension List – takes you back to the list of dimensions registered for this job.
• Up to Parent - for combined dimensions. Moves focus up to the parent dimension.
• Uncombine - for combined dimensions. Extracts the selected dimension out of the combined dimension, allowing separate weighting and processing (go to Uncombining a Dimension for more information).
• Dimension Name – the name of the selected dimension.
• Collapse Type – appears below Dimension Name in the event a panel does not contain enough suitable panelists to fill 'requested' targets in particular segments. This property controls how shortages in segments are to be handled. The options are:
• Nominal collapsing – The targets for the segments with shortages are reduced to correspond to the maximum available in the sample population. The excess (un-allocated) target is distributed proportionately (according to segment targets) to the other segments within the dimension.
• Ordinal collapsing – in shortage scenarios, the total target is kept. The weight targets for the segments with shortages are reduced to correspond to the maximum available in the sample population. The excess (un-allocated) target is distributed to the next segment within the dimension.
• Non-collapsing adjust targets – the total target is reduced by the amount the segment is short. To reach the original target outgo for the other segments, the weight targets for all the segments within this dimension are recalculated to suit the adjusted total targe.
• Non-collapsing keep targets – the total target is reduced by the amount needed to accommodate the original weight target for a segment with shortage. All segment weight targets are kept.

Note: The number of dimensions and the number of segments within the dimensions will significantly affect the time taken to run the sample job. Depending on the composition of the panel, it may also become impossible to fill all cells in the matrix. For extreme matrix definitions with many dimensions and/or large lists of segments within the dimensions, the sample extract may become impossible to perform.

Select two or more dimensions in the list and click the Combine Dimensions button to combine the selected dimensions (go to Combined Dimensions for more information).

Nominal Collapsing

Nominal data is a form of categorical data where the order of the categories is not significant. Examples include industry ("Accommodation, Cafes and Restaurants", "Agriculture, Forestry and Fishing", "Communication Services" ,...), gender, ethnicity etc. In shortage scenarios, the target sample outgo is kept. The segment target of a segment with shortage is reduced to correspond to the maximum available the in sample population. The excess target is distributed proportionately (according to the segment targets) to the other segments within the dimension.

For example: Assume we have asked for a sample of 40000 in total, spread out evenly across the age groups with 10% (4000) taken from each group - . In this case, six of the age groups have shortages, but we see that the total number of panelists requested is kept by filling up from the four segments where we have an excess. Note that the missing panelists are drawn as evenly as possible between these four segments (1413,1412, 861, 1413).

Figure 3 - Example of the results when Collapse Type is set to Nominal Collapsing

Ordinal Collapsing

For ordinal data, the numbers assigned to the objects represent the rank order (1st, 2nd, 3rd etc.) of the entities measured. For this type of data, comparisons of 'greater than' and 'less than' can be made. In shortage scenarios, target sample outgo is kept. The segment target of a segment with shortage is reduced to correspond to the maximum available in the sample population. The excess target (the difference between the requested target and the actual quantity available) is distributed to the next segment within the dimension. If the last segment has a shortage, the redistribution will be done in reverse order - the next to last segment will be given the excess target.

For example: Assume we asked for a sample of 40000 in total, spread out evenly across the age groups with 10% (4000) taken from each group. We have a lot of shortages, and if we look at the figures we see that the shortages for age groups 15-19 and 20-24 are compensated for in the 25-34 segment. The next shortages are in segments 6,7,8 and 9. These shortage are partially compensated for in the following segment (10). However we don't have enough panelists in segment 10 to fully compensate, so then the order is reversed. This means we try to compensate in segment 9, then segment 8 etc. but these are already used up. As many as possible are taken from segment 5 but there are still not enough, so the remainder are taken from segment 4 - .

Figure 4 - Example of the results when Collapse Type is set to Ordinal Collapsing

Non-Collapsing Adjust Targets

Target sample outgo is reduced by the amount the segment is short. In order to reach the original target outgo for the other segments, the segment targets for all the segments within this dimension are recalculated to suit the adjusted sample outgo.

For example: Assume we have asked for 40000 panelists, divided evenly as 50% females and 50% males, which should give us 20000 of each. We have more than enough male panelists available (27620), however we have only 18122 females, giving us a shortage of 1878 females - . The total target is therefore reduced from 40000 to 38122 (40000-1878=38122). The weight ratio for females is recalculated (18122/38122=0,4754 or 47.5%), and the weight ratio for males is also adjusted accordingly (20000/38122=0.5246 or 52.5%).

Figure 5 - Example of the results when Collapse Type is set to Non-Collapsing Adjust Targets

Non-Collapsing Keep Targets

Target sample outgo is reduced by the amount needed to accommodate the original segment target for a segment with shortage. All segment targets are kept.

For example: Assume we are asking for 40000 panelists, divided evenly as 50% females and 50% males. In this case we have more than enough (27620) male panelists, but we have only 18122 females. This gives us a shortage of 1878 females. To keep the 50/50 weight-ratio the male target is also reduced by 1878, so the final target sample outgo is therefore reduced to 36244 - .

Figure 6 - Example of the results when Collapse Type is set to Non-Collapsing Keep Targets