Hi all,
I would like to know if the Periods conversion from weekly to monthly can correctly work in analysis,
Considering the fact that some months will start in the middle of the week.
I have been searching the doc, but haven’t found some useful insights.
1 Like
Hi @kose ,
If data is collected weekly but the aggregated to months, a numerical factor is applied to the data in order to approximate the value for the number of days which fall in all of the weeks. As you noted, weeks may overlap months. So, if three days of a certain week was in one month, and four days were in another month, then a factor of 3/7 would be applied to the first month and then 4/7 to the second month.
So, it really depends on what you mean by “correctly work”. It is simply not possible to exactly aggregate weekly data to monthly data (or quarterly or yearly). It is however important to understand the approximation factor which is used to linearly approximate the contribution of weekly data to monthly totals.
Hope this helps! Best regards,
Jason
2 Likes
Thanks, @jason for the quick insights and noting the approximation.
Actually, in one of the use cases, we were implementing a weekly data collection which will also need to aggregate into monthly. So the period’s correlation and approximation become a big issue. If there’s work on exactly aggregating weekly to monthly without some data being left out from some days, it would be a game changer to our use case.
In this case, you should consider to use daily data capture. You would then not need to approximate anything, and would always be able to aggregate data into either weeks or months.
Its simply not possible really to do this any other way, since weeks simply never add up directly to months. However, capturing the data each day would solve this problem.
2 Likes