CLI reference

tsdive switchback analyze

Report the difference between settings A and B under a plan.

The guide walks a trial with it.

Usage

usage: tsdive switchback analyze [-h] --plan PLAN.json --target TARGET
                                 [--covariate TAG] [--json] [--no-color]
                                 parquet [parquet ...]

Options

Argument	Value	Default	Description
parquet ...		required	archives holding the target and every covariate
--plan	PLAN.json	required	plan file written by tsdive switchback plan
--target	TARGET	required	tag to analyse, as source:point or a point id
--covariate	TAG	[]	tag for the adjusted estimate, declared before the analysis; repeat for more
--json			print one JSON object and nothing else
--no-color			plain text, no ANSI colour (NO_COLOR does the same)

Example

$ tsdive switchback analyze data/switchback_demo/*.parquet \
    --plan data/switchback_demo/plan.json --target TI201.PV \
    --covariate FI200.PV --covariate TT001.PV
demo:TI201.PV   B - A +0.6024 degrees Celsius   p 0.154

plan      digest 66da65ede04f verified   seed 7
schedule  2024-06-03 00:00:00Z -> 2024-06-04 00:00:00Z   (1 d)
blocks    24 of 1 h   A 12   B 12   washout 15 min
design    over 10^6 balanced assignments   1000 drawn   smallest p 0.000999
units     degC -> degrees Celsius
quality   GOOD 1.000   censored unknown

Difference in means  (B - A over the kept samples)
  estimate  +0.6024 degrees Celsius   p 0.154
  95%       [-0.2223, +1.500]
  kept      A 540   B 540   per block 45 to 45

Adjusted  (OLS on 2 covariates)
  demo:FI200.PV, demo:TT001.PV
  estimate  +0.2847 degrees Celsius   p 0.000999
  95%       [+0.2214, +0.3444]
  kept      A 540   B 540   per block 45 to 45

Assumptions
  difference between settings A and B under the declared random schedule, by
  intended assignment; holds if the schedule was followed and carryover ended
  within the washout