Select to view content in your preferred language

Space Time Cube with normalized variable

850
3
Jump to solution
05-31-2018 08:44 AM
DanEvans
Frequent Contributor

I'm looking at analysing numbers of gas service pipe failures to find emerging hotspots. Because there could be different numbers of properties with gas meters in whichever polygon aggregation I use (postcodes, hexagons etc.) I think that ideally the data should be normalised, i.e. failure rate = number of failures / number of gas meters.

Is there a way to create a space time cube that will calculate this failure rate rather than just use the non-normalised count of failures? Or is there some workaround to get the same result to use with the emerging hot spot analysis tool?

0 Kudos
1 Solution

Accepted Solutions
DanPatterson_Retired
MVP Emeritus

I don't like the idea of normalizing that type of case.  I know why you want to do it, but it creates 'fake' hotspot.

Perhaps you can show what you are working with.  I would also make one suggestion ... map the areas with an explicit class where gas meters just don't exist at all.  Anything that you map on top of that will make it quite clear that those areas are out of bounds.  Maybe your 1 failure out of 2 meters will become more obvious.  I often see people mapping 'what' they have without explicitly delineating where 'what' can't exist

View solution in original post

0 Kudos
3 Replies
DanPatterson_Retired
MVP Emeritus

I don't like the idea of normalizing that type of case.  I know why you want to do it, but it creates 'fake' hotspot.

Perhaps you can show what you are working with.  I would also make one suggestion ... map the areas with an explicit class where gas meters just don't exist at all.  Anything that you map on top of that will make it quite clear that those areas are out of bounds.  Maybe your 1 failure out of 2 meters will become more obvious.  I often see people mapping 'what' they have without explicitly delineating where 'what' can't exist

0 Kudos
DanEvans
Frequent Contributor

Thanks for replying.

Below is an example of the data I'm working with. Each dot is at a centroid of a postcode unit, and represents one service pipe failure at that unit (so there are stacks of dots where there's been more than one failure in that postcode).

I've also done a standard hotspot analysis with the failures summarised so that there is only one dot at each location with a count of the total number of failures over a five year period.

I could instead use polygons of each postcode unit and use those with no gas meters as the locations where failures cannot exist, if that sounds sensible to you?

0 Kudos
DanPatterson_Retired
MVP Emeritus

last sentence yes.... because your map... even if hotspot analysis might make someone go whoa! pretty significant.

blot out the 'Void' and you might want to consider just focusing in on the major communities at an appropriate zoom level.

Nothing will be gained by showing an overview map IMHO

0 Kudos