New Approaches to Software for Statistical Processing Joint seminar of the RSSSCS and the ASC Issues

1
New Approaches to Software for Statistical
ProcessingJoint seminar of the RSS-SCS and the
ASCIssues in metadata standards aiming at the
re-use of questionnairesexperiences from IQML,
IT et al.Hans Rutjes Jannes HartkampDESAN
Research Solutions
2
DESAN

• activities
• data collection processing (quantitative)
• analysis reporting
• RD
• methods
• telephone
• paper
• Internet
• face-to-face
• mixed mode

3
Data exchange history

• not so long ago data could not move (punch cards
  could)
• now data exchange is hardly a problem
• metadata exchange is an issue
• CAI metadata exchange is a problem

4
CAI metadata currently

• many different CAI-packages in use
• re-use / sharing of questionnaire scripts and
  data collection metadata problematic
• -gt
• collective benefits of standards for
  questionnaire definitions and data collection
  metadata acknowledged

5
On standard spread

• Standards can emerge in two ways
• A. authoritative body (e.g. )
• B. self-strengthening process (e.g. English)
• Spread -gtattractiveness -gt spread -gt...
• Standards as hyper-collective goods ?
• collective action
• collective benefits

6
CAI md standards currently

• no authoritative body
• no dominant player
• no shared belief in emerging standard
• collective benefits of common standard
• DDI, Dublin Core not rich enough
• several standardising initiatives

7
Standardising strategies

• A. Existing software -gt standard
• advantages (e.g. SPSS)
• software available
• some market share
• B. Multiple packages -gt standard
• (e.g, Triple-S)
• C. New standard -gt software
• advantages (e.g. IQML idea)
• collective product
• easier to optimise as a standard

8
Standard requirements

• standard should be
• rich enough to be useful
• simple enough to be implementable
• standard needs belief to spread
• adoption must be cost-effective

9
Some technical issues

• routing
• (if ?2gt?1 goto ?5)
• separate items or sub-items of questions ?
• dynamic questions and answer cat
• (How long has ?3 lived in ?7 ?)
• missing values
• data collection metadata
• (types of non-response, interview datetime etc)
• flat or hierarchical data?

10
IQML experience (1)

• EU IST FP5 2000-2003
• 5 modules
• (QDT, MR, SAT, QPT, DIT)
• 4 main developers
• (Dimension, UEdin, DESAN, Comfact)
• 2 strategies
• standard -gt software (questionnaires)
• software -gt standard (forms)
• 3 resulting DTDs

11
IQML experience (2)

• theory
• practice

IQML XML
model
QDT
QPT/DIT
MR
SAT
IQML XML
XML4DR
model
RDRMES
QDT
QPT/DIT
MR
SAT
12
IQML experience (3)

• main CAI innovations
• XML based
• distributed
• re-use of questionnaire items, e.g.
• ltTextRefgtIQML6481lt/TextRefgt
• ltText textId"IQML6481" textType"instruction"gt
  ltTextStringgtWhat is your sex?lt/TextStringgt
  lt/Textgt
• concept seems sound, functionality limited

13
IT experience (1)

• IT
• full-fledged generic CAI software developed by
  DESANs sister company Interview Technology
• paper, telephone, web, face-to-face, self, mixed
• used by gt100 data collectors world wide
• data export to e.g. SPSS, dBase, Lotus, Quantime,
  SAS, Triple-S
• (see Feb 2004 issue of Research)

14
IT experience (2)

• IT -gt TripleS export
• relatively easy to implement, but
• no user/system missing distinction in Triple-S
• 0 not allowed as category in Triple-S (negative
  nr is)
• sequential category numbering in Triple-S
• no MR for numeric variables in Triple-S
• TripleS only variables and data (no routing
  etc.), so IT -gt TripleS -gt IT would mean huge
  loss
• low costs, low benefits ?

15
IT experience (3)

• IT -gt SPSS MR Data Model retrieval
• implementation just started
• on-line data retrieval
• model seems complex, but well-documented
• questionnaire dynamics (skips, loops, random...)??

16
Way forward

• standards can start with basic elements and be
  extended and diversified later
• 10 of standard may suffice for 90 of use

17
Ideal?

• consortium of multiple key CAI-players
• index and analysis of functionality/issues
• questionnaire dynamics
• validation
• mixed mode
• collection metadata
• model development
• interface development

18
CODACMOS

• EU IST FP5 Cluster 2002-2004
• focus on data collection
• partners
• NSIs of Italy, Finland, Norway, Greece
• universities of Edinburgh, Patras, Athens
• others (DESAN et al.)
• issues for future research
• attention for non-technical issues

19
Info

• www.desan.nl
• www.epros.ed.ac.uk/iqml
• www.InterviewTechnology.com
• www.triple-s.org
• www.codacmos.eu.org
• www.research-live.com (IT review Feb 04)
• rutjes_at_desan.nl / hartkamp_at_desan.nl