Skip to main content
cancel
Showing results for 
Search instead for 
Did you mean: 
Anonymous
Not applicable

Problems Implementing MultiVariate Measures

I have a standard data model that takes all observations of scientific measurement values in a single large table. Here is a subset:

    quantity AS [QUANTITY NAME],

    unit AS [UNIT NAME],

    measurement_value AS [MEASUREMENT VALUE],

Now, quantity name may include many dimensional values:

electric flux (flux of displacement)

quantity of electricity

angular frequency

gyromagnetic coefficient

angular repetency

angular wavenumber

attenuation coefficient

gyromagnetic ratio

The quantity name has a corresponding one or more units. This may include, for example:

ampere second

ampere second

hertz

ampere square metre per joule second

radian per metre

radian per metre

reciprocal metre

ampere square metre per joule second

For each measurement, and there may be billions of rows, the actual measured numerical value is stored in [MEASUREMENT VALUE]  - for instance, a row may look like:

[QUANTITY NAME][UNIT NAME][MEASUREMENT VALUE]
electric flux (flux of displacement)ampere second1.232324
quantity of electricityampere second.0000999988888
angular repetencyradian per metre121.56434566

Now, while this is wildly efficient, it raises an issue when using Sense. I have many dimensions, but only one measurement to plot. Hence, its not obvious how to plot or compare multiple measurements (multivariate charting and reporting).

Hence, I want to chart  quantity of electricity as expressed as  ampere second  and also show the relationship to  electric flux (flux of displacement) as expressed as  ampere second. In a perfect, though much less efficient world, I would have a single column, not row, for each combination of quantity and unit name, and the rows would be the values. Then I would have very clear measurements and Sense would be smiling all the way to the UI. However, I have up to 10,000 different classifications of measurements, and no database is going to store some 10,000 columns with billions of rows.

So, the question is: how do I plot a multivariate chart when the data is in this format, without overburdening the user with manual work?

3 Replies
ahaahaaha
Partner - Master
Partner - Master

Hi Chris,

If I understand you correctly, then you can  anything not to change  in the accumulation of your data in one large table. Further when constructing the necessary charts, you can apply selections, bookmarks and use in diagrams of Set Analisys expressions to obtain the desired result.

Regards,

Andrey

Anonymous
Not applicable
Author

I should clarify. So the user interface should be simple - meaning the data needs to be easy to consume (aka: should not need to devolve to set analysis to do multivariate drag and drop). Its plausible to change how data is injected, but there are bigger issues (10,000 columns...), so not an immediate solution there.

Given this, then the question is if at ingestion Sense can break out the rows into columns in the QVD, and hence resolve the core issue.

Anyone have a thought? Kind of a pivot import (rows to columns)? So from the above to the below:

[Type][parentID][quantity of electricity | ampere second][electric flux (flux of displacement) | ampere second][angular repetency | radian per metre]
measurementx.0000999988888
measurementz1.232324
measurementy121.56434566
brunobertels
Master
Master

Hi Chris

have a look in the help Qlick Sense Help section for Generic load :

https://help.qlik.com/en-US/sense/3.2/Subsystems/Hub/Content/DataSource/generic-databases.htm

So have a look here also

https://community.qlik.com/docs/DOC-3020

This document helps in understanding Generic Load concept which pivots/transpose given data.


hope it helps