Drifts 10: barometer error estimation

[DavorJ]

The following is based on #51. Note the slide that describes the underlying reasoning:

Let us take barometer BAOL001X_D0939 as an example. Analysis #51 revealed the following pattern when compared with all other barometers:

Let's extract the red curve which is the median of all the differences between measurements of BAOL001X_D0939 and all other barometers, given a certain time t.

We can make a histogram and a boxplot of the red curve:

which show that the red curve falls nicely between -10 and 10 which makes the difference between whiskers around 16 cmH2O. You can find BAOL001X_D0939 at the 18th place from the top of the following graph, which is a comparison of all the barometers, following the same procedure, ordered by the descending whiskers difference "WISK".

Here is also a CSV export with extra fields such as IQR and 95 % interval. Note also that I only consider the red curve at a certain time t if there are at least 10 other barometers available at that time.

Does this say anything about the error that BAOL001X_D0939 is making?

Yes and No.

"No", because the error of the barometer is , while our red curve is based on

in which is hidden.

"Yes" because we are comparing all the barometers on the same level, so their variance is in some way proportial to their true , which allows for comparison.

Can we make an estimate of ?

Yes we can, To be discussed...