From Data to Paper via Stata

1
From Data to Paper via Stata!
ESRC Funded Researcher Development
Initiative Project Grant Course 2 From Data
to Paper (via Stata)

• Tim Croudace
• and Jon Heron
• Jon works in Bristol too -)

2
From Data to Paper via Stata!
ESRC Funded Researcher Development
Initiative APPLIED PSYCHOMETRICS Project Grant
Course 2 From Data to Paper (via Stata)

• Tim Croudace
• and Jon Heron
• Jon works in Bristol too -)

3
From Data to Paper (via Stata)
The origins and genesis of (new) psychometric
publications (for you)
Newnham College, May 2008

• Tim Croudace tjc39_at_cam.ac.uk
• Department of Psychiatry
• 1- Senior Lecturer (Psychometric Epidemiology)
• 2- Dept of Health Career Scientist (Public
  Health)
• 3- Director of Research The Psychometrics
  Centre, Cambridge

4
From Data to Paper via Stata!
5
Whats new in psychometrics (1)

• Locally- quite a lot!
• Places
• New project educational training project
• recently underway
• Helping more people understand the potential for,
  and how to do psychometric analyses (better than
  they currently do)
• ESRC Researcher Development Initiative (RDI)
  Project Grant
• Applied Psychometrics with Profs Susan
  Golombok John Rust

6
http//psychometrics.sps.cam.ac.uk
7
Whats new in psychometrics (2)

• The frontiers of modern multivariate analysis
  have moved quite a long way dont get left
  behind
• There are many new procedures in modern
  statistical computing environments that can
  improve your research
• e.g. the recent explosion of new psychometric
  routines for quantitative modellers in modern
  statistical computing packages
• such as Mplus, Splus, R and Stata
• see. Journal of Statistical Software, Special
  Volume Psychometrics in R
• See The Stata Journal, www.gllamm.org or check
  out whats new in SAS!
• In the same way that regression has moved into
  multilevel, so has factor analysis, structural
  equation modelling and IRT
• Many of these innovations are due to the power
  and flexibility of ML estimation
• Old favourites are still available and still very
  useful, just more powerful once extended e.g.
  Latent class analysis
• standard model still available but extended to
  include random effects
• now applicable to mixed measurement level
  outcomes
• and adapted for longitudinal (panel data)
• and multilevel settings (family data)

8
From Data to Paper via Stata!
9
From Data to Paper via Stata!

• Introduction to data preparation
  considerations/advice for psychometric scaling /
  psychometric statistical analyses
• Demonstration (not practicals) of main commands
• in Stata for convenience (to us)
• Helping you to structure your data-based
  analytical activities before writing up papers
• Parallel Development of Course Website
• Logfile of all demonstrations run during the
  course
• Linkage to example publications
• Linkage to websites for other software(s)

10
From Data to Paper via Stata!

• Introduction to data preparation considerations
  for psychometric analyses
• Data collection
• Data entry
• Data validation
• Data transformation and recoding
• Data description and representation
• Data display and (uni-)dimensionality
• Psychometric statistics and scaling models
• Simple/Summary, Non-parametric, Parametric

11
Data

• Data collection
• How you do it is down to you, depends on what you
  are doing
• Data entry
• Once again down to you, or someone else, unless
  you beg and borrow all your data from ongoing
  studies (like Jon and I do! not true actually
  -) )
• Data validation
• Very important only allowable responses should
  be analyses, control over labelling and missing
  values very important too
• Most second hand datasets given to me contain gt1
  error

12
Data

• Transformation and Recoding
• GNIROCS Reverse scoring (geddit?!)
• When and Why Question answer combinations
• Alternate scoring methods / variable recoding
• Categorical variable collapsing e.g binary
  recoding
• Data description and representation
• Summary statistics, distributions, displays
• Data dimensionality and display
• Eigen values, PCA Factor Polychoric PCA

13
Some Sage advice

• 3 books
• Scaling Procedures Issues and Applications
• R.G. Netemeyer, W.O. Bearden and S. Sharma
• Fundamentals of Item Response Theory (IRT)
• R.K. Hambleton, H. Swaminathan and H.J. Rogers
• Introduction to Nonparametric IRT Mokken
  Scaling
• K. Sijtsma and I.W. Molenaar
• Test Theory A unified approach
• R.P. McDonald.
• Comprehensive Psychometric Overview
• Linear Factor Analysis
• Categorical Data
• (Binary Factor AnalysisItem Response Theory)
• Both understood through the Common Factor Model

14
Recommended books on general multivariate analysis

• The Analysis and Interpretation of Multivariate
  Data for Social Scientists AIMDSS
• David.J. Bartholomew et al. LSE teaching team
• 2nd edition due out this/next month
• Multivariate analysis techniques in social
  science research From Problem to Data
• Jacques Tacq

15
(No Transcript)
16
What we have in common? - multivariate data
17
(No Transcript)
18
GENESIS study composite index a dimensional
phenotype
19
Data of mixed measurement level binary/ordinal