Weighting prepare and calculate
How to calculate the weights for respondent weighting using targets using QPSMR Companion. See Weighting respondents for an overview of rim/target weighting.
Prepare and check data
Before the calculations can begin you need to prepare the data with [Calculate] [Prepare and check data]. This checks the data for all the targets and sets all the weights to 1.0.
If any targets have no usable data you should set these targets to “Ignore”. You may have to redistribute the target to other responses so that the total remains the same (usually 100%).
If the data has been weighted earlier you can choose to start with those weights rather than starting every record with a weight of 1.0. If any records have been added since the weights were calculated you will need to manually set their weights to 1.0. This option could be used for “two stage” weighting.
[Calculate] [Calculate weights] will start a number of iterations (passes) over the data adjusting the individual weights to try to reach the targets. This process will continue until all the targets have been reached or no further progress can made.
If the number of iterations is excessive, then there is probably a problem with the targets or the data. You can stop the iterations at any time and inspect the summary (top of window) or the individual data records (bottom of window) to see where the problem lies.
In normal market research the weighted totals do not need to be exactly the same as the targets. A total that is very close is good enough. The default precision is 8 significant digits, meaning that the first 8 digits of the weighted total will be the same the target. Companion will cease iterations when every total is this close to its target.
If your weighted totals are very large you may want increase the precision. You can do this after calculations have been done and then use [Calculate] again without [Prepare and check data] beforehand. You only need to [Prepare and check data] once unless you have changed the targets, the weight question, or the data file. The higher the precision the more iterations will be needed and the longer it will take to calculate the weights.
The individual weights are stored in the float question and some precision may be lost if insufficient data locations have been allocated to the question. To see the values that will actually be stored use [Calculate] [Show stored weights].
You can choose to set minimum and maximum weights. If the weight for a record reaches the limit set, then the weight for this record will not be allowed to go outside the limit. This may result in not being able to reach all the targets set.
If you close the Weight data window without saving the data you will be prompted to do so by Companion.
Important: The newly calculated weights are not stored in the data until you save the data file. Before you do this you should check and save any reports that you might need.