Question: How do I calculate error for averages?

wu_ming2 is asking a question about purpleair
Follow this topic

by wu_ming2 | February 02, 2019 03:27 | #18243

I just installed the device indoor to check PM2.5 air quality in our flat and how the new air purifier works. After a while I am planning to move it outdoor.

As mentioned elsewhere I developed a micro service to notify me when readings cross a threshold. I have those values expressed as a range since we have the error range for the sensor.

PA-II comes with two PMS 5003 laser counters to estimate air particulate concentration. Specifications sheet among other things cite a maximum consistency error for PM2.5. This tells me the internal precision w/o reference to a true or known value. But what kind of distribution is assumed for this precision estimate? And what is the significance? As example I would have expected to read +-10% with a significance of 2.5 standard deviations for a normal distribution.

Without this piece of information I wouldn't know what probably range to assign and how to calculate the precision for averages between the two sensors and for different timeframes. Am I overthinking it and +-10% is just fine enough? Thanks.


I'm going to share this on the Public Lab Twitter account to see if we can get some more input from others

Reply to this comment...

Hi @wu_ming2 - sorry we weren't able to connect you with some resources on this. This comment from @BrandonFeenstra may have some useful information:

If you would like to learn more about Air Quality Sensors and their performance against regulatory instrumentation, I would recommend a visit to the South Coast Air Quality Management Districts's (SCAQMD) Air Quality Sensor Performance Evaluation Center (AQ-SPEC) web page. The purple air sensor has been evaluated in both the field and the laboratory. Program's website: PA-II field report: PA-II laboratory report:

Ah, and i see you found @guolivar's post here:

Thank you!

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Log in to comment