Home/ Questions/Q 7010511
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-27T22:00:00+00:00

Here’s what I really mean: I am using test scripts on groups of devices

  • 0

Here’s what I really mean:

I am using test scripts on groups of devices looking different ‘soft-failure’ type conditions.
For example: Test a group of radios for signal strength on different frequencies.
I want to find the test condition (frequency in my example) that yields the best result
for all devices (radios).

I’m having trouble clearly identifying a mathematical model for this.
Please tell me I’m overthinking this and there’s some obvious method I’m missing!

At first glance, you might think I just want the average, but that’s not really any good.
7 out of 8 radios might have great signal strength on one frequency, but the last radio may not work well. This scenario may have a good average value, but doesn’t measure the distribution of the values. The median value may be closer to what I need, but it still doesn’t work.

The end goal is a metric that can be used to measure a group’s results for how well the group’s combined performance compares on each setting- Taking into account the distribution of the values. Does anyone know if this be done without using thresholds?

Random data example for 3 different frequencies (Higher values are better):

             F1       F2     F3
 Radio 1    -55      -65    -40
 Radio 2    -60      -66    -99
 Radio 3    -65      -67    -70
 Radio 4    -70      -68    -80
 Radio 5    -99      -69    -50
 Radio 6    -50      -68    -60
 Radio 7    -65      -69    -60  
 Radio 8    -70      -68    -70

 Median     -65      -68    -65
 Average    -66.75   -67.5  -66.13

Based on the condition that higher values are better, F2 is the best choice here, despite it having a poorer Median and Average.

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    Use standard deviation? I think you’ll still need some kind of threshold – for example, I imagine there is a set where they are all tightly distributed around a poor value (say -72) which wouldn’t be as preferable to most working well and one working badly. If there is some strict bail out value (like any radio worse than 90) you could test for that explicitly.

    If all you care about is uniformity though, just test standard deviation.

    • 0