Reading the compare report¶

tsdive compare measures what changed between two periods over every tag of a unit. This is the report on the demo flow and temperature, before 20:00 to 23:00, after 04:00 to 06:00, then every line.

$ tsdive compare data/demo/fic101_demo.parquet data/demo/tic101_demo.parquet \
    --before 2024-03-30T20:00:00Z/2024-03-30T23:00:00Z \
    --after 2024-03-31T04:00:00Z/2024-03-31T06:00:00Z
demo  2 tags   refused 0   pairs 1 of 1 clearing

before    2024-03-30 20:00:00Z -> 23:00:00Z   (3 h)
after     2024-03-31 04:00:00Z -> 06:00:00Z   (2 h)
grid      rate 60 s (declared)   coverage 1.000 -> 1.000

Tags that changed
  tag        quality   sigma  spread  flagged  changed at
  TIC101.PV  ok         +0.8    x6.0      62%  2024-03-31 04:18:00Z
  FIC101.PV  ok         +0.4    x0.5       0%  2024-03-31 04:00:00Z

Pairs that decoupled  (Pearson on first differences, 95% block bootstrap)
  pair                   before   after   delta  interval
  FIC101.PV ~ TIC101.PV    0.69    0.17   -0.53  [-0.76, -0.26]

Joint structure  (PCA fitted on before, 2 of 2 tags aligned)
  components 1 of 2   explained 0.9767 -> 0.5129 on after
  rows 121   T2 breaches 70   SPE breaches 108
  SPE contributors   TIC101.PV 50%   FIC101.PV 50%

Headline and periods¶

demo 2 tags refused 0 pairs 1 of 1 clearing: the source, the tags compared, the tags whose read was refused, and the pairs whose interval excludes 0.
before and after: the two periods and their lengths.
grid rate 60 s (declared) coverage 1.000 -> 1.000: the rate the pair and joint tables align on, and the share of grid cells filled in each period.

A refused table prints its reason under the headline, such as a missing sample_rate_s.

Tags that changed¶

Every tag is listed, ranked by the size of its level shift.

TIC101.PV ok +0.8 x6.0 62% 2024-03-31 04:18:00Z: after-period quality is ok; the median rose 0.8 before-period sigmas; the MAD is six times the before MAD; a screen built on the before period flags 62% of the after samples; and the first change point after 04:00 is at 04:18.
FIC101.PV ok +0.4 x0.5 0% 2024-03-31 04:00:00Z: the flow moved 0.4 sigma, got quieter, and no sample leaves the before limits. It changed little.

Pairs that decoupled¶

FIC101.PV ~ TIC101.PV 0.69 0.17 -0.53 [-0.76, -0.26]: the minute-to-minute correlation of flow and temperature was 0.69 before and 0.17 after. The drop of 0.53 has a 95% interval from -0.76 to -0.26, which excludes 0, so the pair clears.

Joint structure¶

components 1 of 2 explained 0.9767 -> 0.5129 on after: the PCA fitted on the before period keeps one component, which carries 98% of the before variance and 51% of the after variance.
rows 121 T2 breaches 70 SPE breaches 108: after rows, and those above the before period's T2 and SPE limits.
SPE contributors TIC101.PV 50% FIC101.PV 50%: with two tags and one component the residual falls on both tags equally, so the shares name neither. The tag table names the temperature.
When the model keeps every component, no residual is left: the breach line reads SPE NOT ASSESSED and the next line gives the reason in place of the contributors.

What the demo says¶

The temperature got noisier and stopped following the flow from about 04:00; the flow barely changed. The report does not say why. A fouled thermowell, a loose sensor and a new heat source give the same tables. Compare's three tables explains each column.

What to do next¶

Start with the top row of the tag table and its changed at time.
A clearing pair: check the process link between the two tags at that time.
The tables as data: --json, or CompareAnalysis.to_dict() in Python; CompareAnalysis.frame holds the tag table.