Forum Discussion
case weighting in data
Hello,
as I am working in market research, our data which is always a sample from the population, is often weighted (nearly always). In statistical programs such as SPSS, one can define a variable which holds the case weights and activate that variable as a weighting variable (i.e. for cases overrepresentated in a sample, case weights are >1, whereas for cases underrepresented in a sample, case weights are <1). Is there a way to achieve that in Power BI?
Because, frankly, if there is no way for that, how could we ever use (raw) market research data in Power BI, because such data is very often weighted.
Any ideas would be helpful - thanks in advance,
Daniel
Okay, after a few trials and errors, and googling the DAX formulas, I might have worked it out. so in my case the formula I needed was, for displaying the average of a scale value (index value, ranging from 0 to 100):
MyWeightedMeasure = SUMX('TableName'; 'TableName'[WeightColumnName] * 'TableName'[ColumnName]) / SUM('TableName'[WeightColumnName])
I realized that this "workaround" was only necessary for displaying such an average value. For percentages, i.e. frequencies of groups that are represented by different names in one column, such as "male/female", you only need to put your column in the legend field, and your weight column in the value-field (I'm in the german version, I hope I translated the words correctly). But be careful, I think that only works right if your weights sum up to the total n of cases in your table.
9 Replies
What have you tried so far? This is a bit outside my "sweet spot", but I would recommend looking at R to do some of this for you. There is R integration for PowerBI and I think that is where you will get this type of functionality. Use this Link to get started.
Thanks for your suggestion. In R, weighting cases seems to be a bit more difficult than it is in SPSS, where you simply type "weight by var" and you're done. Also, if Power BI doesn't has weighting implemented, then I could also switch completely to R, I think. But my experience with both of the programs is very little...
Bumping this as I have the same question - Power BI does not look like it would be useful for analyzing/visualizing survey data if there's no option to set a case weight! I realize I could switch to R, but I was hoping that Power BI would allow me to produce visualizations quickly, and having to interface with R will make it slower than SPSS, Stata, etc.
