For fellow nerds, here are more details on our data and methods:
Methods: Anna turned the purifier on the highest setting before she went to bed at night and turned it off after she woke up. To measure particulates, she used a Dylos particle counter, which measures particles .5 microns and above and 2.5 microns and above per .01 cubic foot. The particle counter took one measurement each hour.
Calculating Effectiveness: Thomas calculated effectiveness as the percentage of particles the purifier removed from the room air. The baseline was the particle count before turning on the purifier. The final count was the average number of particles over the last four hours before Anna woke up. We prefer this over comparisons to outside air because:
2. The baseline room number takes into account how dirty the outdoor air is because indoor particulates go up and down with outside air.
However, the drawback is that outdoor particulates sometimes go up after the test starts (lowering estimates of effectiveness) and sometimes go down (raising estimates). However, these average out over multiple tests, and the results are similar if you look only at days where outdoor particulates were relatively stable.
Room: Anna’s bedroom is 15 m2, located in Chaoyangmen, Beijing. The doors and windows were closed while Anna slept, but she opened the door at various times in the very beginning of the test before she went to sleep. Although the doors and windows were closed, the apartment isn’t new, and the seal isn’t great.
For more details on the methods, see the end of Thomas’s earlier post. All methods were identical to those earlier tests.
Filter Life: After all the tests were done, the IQ Air said the pre-filter had 1,931 hours of use left, the carbon filter had 3,077 hours, and the HEPA had 1,910 hours. Thus, the filter was in its prime.
Outlier: Of the 11 tests, one day was a strong outlier. On May 22nd, the IQ Air got only 68% of the .5 micron particles. Normally when we see poor results, it’s because the outdoor air got a lot dirtier during the night. But on May 22nd, the outdoor concentration fluctuated between 74 and 110 micrograms, which isn’t out of the ordinary.
Because we couldn’t see any reason the results were poor that night, we left the data in. If we redo the analysis with that day included, the results are only slightly better: 93.0% of .5 micron particles and 96.5% of the 2.5 micron particles. That would put the IQ Air about equal with the Philips and still 4% below the Cannon.
For fellow nerds, here are some more details on how the tests were conducted:
The test above was done starting at 11:30 pm on 6/16/2013, when the outside AQI in Beijing was 230 according to the US embassy’s AQI Twitter feed. (The outside air improved the next day, but results were similar on a later test where AQI actually went up slightly from 195 to 202 during the test.)
According to comparisons of our particle counter’s tests of outside to US embassy AQIs, an AQI of 230 would convert to about 2,650 on the PM 2.5 count on the reader. (Remember, the particle counter gives the raw number of particles 2.5 micrometers and above per .01 cubic feet. The US embassy takes mg/m3 and converts that to an AQI. Therefore, the raw numbers are different, but they correlate highly. You can think of this as the difference between measuring attendance at a basketball game by the number of people in the stadium versus the total weight of those people. The two numbers are highly correlated, but not identical.)
Thomas did the test in his bedroom with the doors and windows closed. The room is 13.5 m2, with two windows.
The particle counter tends to take 5-10 minutes to get stable readings, so to be conservative, Thomas gave it about an hour:
The spike at the top was when Thomas entered the room to turn the air purifier on and reset the machine, so it may be the dust he kicked up by walking around. A more stable reading for that time would probably be about 230.
The uptrend prior to turning the filter on may have been because Thomas was running a dehumidifier (which itself has some small filters) in the room before the tests.
Regardless, this data suggests that the effect of the filter was not a confound of calibration.