1351686 - Provide sample Quantum experiment data

Status: RESOLVED FIXED

(Cloud Services Graveyard :: Metrics: Pipeline, enhancement, P1)

Description

[Rob Miller [:rmiller]]

We need some sample data that matches (or is at least close to) the shape of what will be generated by the Quantum feature-flipping experiments, which we can use for the development of our distribution viewer data import and front end dashboards. This data set should contain data for at least two experiments, and at least one of those experiments should contain at least three cohorts.

Thank you!!

Comment 1

[bugzilla]

I've dumped the first batch of sample data into s3://telemetry-test-bucket/ssuh/fake_experiment_analysis (and told Rob Hudson about it yesterday.) The histograms are real collapsed histograms from a sample of longitudinal data where clients were randomly assigned into "branches", and the branches are currently: "control", "branch1" and "branch2"

The current schema looks like:
experiment_name                   string                           required
experiment_branch                 string                           required
subgroup (currently os/version)   string                           not required
n (total number of submissions)   long                             required
metric_name                       string                           required
metric_type                       string                           required
Histogram                         map(key[long], (HistogramPoint)) not required
statistics                        array[Statistic]                 not required

HistogramPoint
--------------
pdf                               double                     required
count                             long                       required


Statistic
---------
comparison_branch                 string                     not required
name                              string                     required
value                             double                     required
confidence_low                    double                     not required
confidence_high                   double                     not required
confidence_level                  double                     not required
p_value                           double                     not required


I'll generate a second "experiment" today and add to the same folder

Comment 2

[bugzilla]

Added the second experiment to the folder. Whoo!