Skip to content
GitLab
Explore
Sign in
Register
Primary navigation
Search or go to…
Project
SERL
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Wiki
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Deploy
Releases
Model registry
Monitor
Incidents
Analyze
Value stream analytics
Contributor analytics
Repository analytics
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
Community forum
Contribute to GitLab
Provide feedback
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
SERG
SERL
Commits
8e83bb9d
Commit
8e83bb9d
authored
5 years ago
by
Ben Anderson
Browse files
Options
Downloads
Patches
Plain Diff
added noted; added UKDS safe researcher training notes
parent
31ea0bc2
Branches
Branches containing commit
No related tags found
No related merge requests found
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
notes/2020-05-11-UKDS_safeResearcherTraining.md
+60
-0
60 additions, 0 deletions
notes/2020-05-11-UKDS_safeResearcherTraining.md
notes/README.md
+7
-0
7 additions, 0 deletions
notes/README.md
with
67 additions
and
0 deletions
notes/2020-05-11-UKDS_safeResearcherTraining.md
0 → 100644
+
60
−
0
View file @
8e83bb9d
# UKDS Safe Researcher Training
May 11th 2020
Case Officer - if want to vary research or pursue new idea with same data, scope creep is OK, can be approved, can need a new project.
## SDC in theory
UKDS/ONS: Threshold = 10 (cell counts)
0 can be a problem unless they are logical - e.g. no 16 year olds have degrees in a table or age x quals
Watch for dominance e.g in % ranks etc
Avoid min/max -> unless have 10 cases (rounding?) Do you need to publish this?
Beware secondary disclosure - e.g. subtract 1 table counts from another -> do you get small cell counts?
## SDC & research
Descriptive tables are a pain - inherently high risk
Regression coeffs etc are easier
No automated output checker
Avoid skewed distributions, small numbers, huge outliers, dominant observations, rare events etc
Avoid min/max values unless can prove not identifiable. Same for median (which is on a box plot)
If asked for new details etc always add these to the outputs - never in an email
Raise/notify an issue early
## Output checking
At end of project ask for output which is syntax files & they will send
All sessions are recorded - can be used to resolve issues etc
Always include underlying counts for Figures (unweighted)
Beware summary tables which include all units and then categorised tables which leave out a row/cell with 1 item in it.
Maps - use heat maps -> intensity/prevalence, much safer. Not point maps.
Put yourself in checker's position! Make it easy.
'Good' output:
use https://www.ukdataservice.ac.uk/help/get-in-touch.aspx
-> output request
-
publishable quality output (word & excel tables?)
-
doesn't support knitr etc - does SERL portal?
User Guide: https://www.ukdataservice.ac.uk/media/622685/sa_user_guide.pdf
This diff is collapsed.
Click to expand it.
notes/README.md
0 → 100644
+
7
−
0
View file @
8e83bb9d
# SERL
Repo to support work on the SERL project (https://serl.ac.uk/)
## Notes
From meetings, training etc. Public stuff.
\ No newline at end of file
This diff is collapsed.
Click to expand it.
Preview
0%
Loading
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Save comment
Cancel
Please
register
or
sign in
to comment