Helps to calculate values for intervals by transforming potentially unscaled interval (e.g. 08:22 – 17:34) into a series of scaled intervals (e., 08:30 – 09:00, 09:00 – 09:30 …)
Introduction | To decide whether such a scaled interval is considered to be a interval caused by the original interval is surprisingly tricky. A nice analogy (by Wolfgang Aigner) is a mesh and the question whether each hole is filled or not. This problem can be further refined by two aspects: - When do short intervals fill larger holes E.g. if a customer waits from 8:02 to 8:05 …is this to be counted as a waiting customer in the interval 08:00 – 08:30?
- How many holes of the mesh are filled by a long interval E.g. how many 30' intervals are covered by the interval 08:17 – 17:14? Is 08:00 to 08:30 and 17:00 – 17:30 an interval? Are both considered, one …?
|
Situation | Raster the time interval 2007/01/01 6:00 to 2007/01/01 14:30 into 1-hour-intervals
|
Operation setting | Choose date columns, scaling and calculation method (other examples see below).
![](_scroll_external/attachments/b2f94f23e632/scaling_ex1_100-a.png) |
Result | ![](_scroll_external/attachments/36df2a94eb47/scaling_ex1_100-a_result.png)
|
TIS Project | Confluence Op Scaling.gzip |
Settings
Time periods are set in a specific scale (minutes or hours).
Examples
The time interval 2007/01/01 6:00 to 2006/01/01 14:30 is applied to different calculation methods.
Calculation method | Settings | Result |
---|
See example above: 100% in a 1-hour raster if Convert 0 lines? is not selected | ![](_scroll_external/attachments/b2f94f23e632/scaling_ex1_100-a.png)
| ![](_scroll_external/attachments/36df2a94eb47/scaling_ex1_100-a_result.png)
|
100% in a 1-hour raster if Convert 0 lines? is selected | ![](_scroll_external/attachments/85748dcdc564/scaling_ex1_100-b.png)
| ![](_scroll_external/attachments/4c04453cf1ae/scaling_ex1_100-b_result.png)
|
50% in a 1-hour raster no matter if Convert 0 lines? is selected or not | ![](_scroll_external/attachments/2087fb478cfe/scaling_ex1_50-a.png)
| ![](_scroll_external/attachments/377c209b91ef/scaling_ex1_50-a-result.png)
|
The following time interval, value = 27 is rastered with different scaling methods.
![](_scroll_external/attachments/c592470878df/scaling_ex2_table.png)
Calculation method | Settings | Result |
---|
Proportional time*value | ![](_scroll_external/attachments/03cead438216/scaling_ex2_settings_a.png)
| ![](_scroll_external/attachments/1992fcb19998/scaling_ex2_results_a.png)
Explanation: Scaling interval = 2 minutes (120 seconds) determines the resulting proportion row 1: half of the interval met, therefore 60/120*27=13.5 rows 2 and 3: interval fully met, therefore 120/120*27=27 row 4: half of the interval met, therefore 60/120*27=13.5 |
Proportional time | ![](_scroll_external/attachments/6864b9a75806/scaling_ex2_settings_b.png)
| ![](_scroll_external/attachments/e303b32bf9de/scaling_ex2_results_b.png)
Explanation: Duration of the period in the whole data set (6 minutes = 360 seconds) determines the resulting proportion row 1: 16:13 – 16:14 = 60 seconds, therefore 60/360*27= 4.5 row 2: 16:14 – 16:16 = 120 seconds, therefore 120/360*27= 9 row 3: 16:16 – 16:18 = 120 seconds, therefore 120/360*27= 9 row 4: 16:18 – 16:19 = 60 seconds, therefore 60/360*27 = 4.5 |
Troubleshooting
Problem | Frequent Cause | Solutions |
---|
There are surprisingly high requirement peaks! – How can I check if they are plausible? | | The following steps can help understand these peaks: 1) Check if the result is due to the scaling method: - Do the peaks disappear, if a shorter interval is selected?
- Do they disappear, if 100% is selected (to exclude rounding)?
2) Check if the peak is due to a few individual values: 3) Check the underlying data. In case the peaks are really errors, eliminate the faulty data records or time intervals e.g. with TIS:Zeitbereichsfilter (old Wiki) . |
There are only points in time. What shall I do? | | Use "Calculation" to add e.g. 30 minutes to create a virtual interval and then analyse with >>Count only start time<<. “Normal scaling” would not make sense in this case. |
The calculated staffing level seems incorrect | This may be due to the interval selected. | If in doubt, try using shorter intervals Example: 15 minute scaling can produce strange results for intervals of 10 minutes or less: in case of 50% overlapping e.g. - 00:00 – 00:10 --> staffing level = 1 in 00:00 – 00:15
- 00:20 – 00:30 --> staffing level = 1 in 00:15 – 00:30
- BUT: 00:40 – 00:50 --> staffing level = 0 in 00:30 – 00:40, 00:40 – 00:50, 00:50 – 00:60
- in case of 100% overlapping --> staffing level = 0
|
- A detailed description of how to allocate time data along a time raster (time interval) using different calculation methods can be found under Basics: Scaling and time raster
- This operation can be used to determine how many persons were present in certain time intervals.
- Please also refer to the operation Sum and count
- Please execute Link to calendar only after scaling, otherwise two raster intervals will be created per type of calendar day.