===== Improving the evaluation of therapeutic interventions in multiple sclerosis: The role of new psychometric methods. =====

==== Chapter 1 ====
=== New psychometric methods ===
Rasch measurement vs Item response theory

=== Traditional psychometric methods  ===

  - psychometric -- originally psychology but now broadened
  - psychometric: methods for developing and evaluating rating scales, and for analyzing their data
  
=== Rating scales ===

  - for properties that we can't measure directly e.g. happiness, disability (vs weight)
  - latent traits -- "inferred variable" = aka 'property', 'trait', 'concept', 'aspect'
  - Dichotomous (RMI) vs polytomous (3 or more) 
  - Direct also opposite RMI vs MSIS-29
  
==== Evaluation of a rating scale ====
  
   " extent to which a quantitative conceptualization has been operationalized successfully "
 
  - Traditional 3: reliability, validity, responsiveness
  - Hobart 6: 

  * Data quality
  * scaling (assumptions)
  * targeting
  * reliability
  * validity
  * responsiveness
  
=== Classical Test Theory (Assumptions x 5) & limitations====

  -  Weak true core theory
  - 'weak assumptions leads to weak conclusions'
  
Limitations of traditional psychometric methods
  - ordered counts are not interval measures
  - results for scales are sample dependent (therefore unstable)
  - results for samples are scale dependent
  - missing data
  - SEM
  - scaling

====Chapter 2 New Psychometric Methods    2021-03-13 (1)====

===Overview===

  - Traditionally use Likert summated ratings --> cannot verify data
  - New: theories can be formally and rigorously tested
    - IRT & Rasch: different origins, perspectives and proponents
    - weak theory --> weak conclusions -- we need strong theory (IRT and Rasch)
  
  - Mathematical relations between variables and events enables __checking, prediction, analysis__ (refine model, review data)
  
 
 New methods also concern the relationship between the TRUE and OBSERVABLE score, BUT also
 

   "focus on the UNOBSERVABLE measurement on the underlying trait and the probability of responding to one of the the response categories of a scale item"
   
   Therefore a focus change to ITEM rather than TOTAL SCORE
   
 ====History====
 
   - IRT: Louis Thurstone UC, Frederick Lord - Educational Testing Service
   - Georg Rasch, Copenhagen - Poisson; each Student as an individual
     - Stability: __item locations__ and __person locations__ could be estimated independently of each other
     - Mathematical model given primacy == 'justification for model selection is theoretical evidence of its suitability'
     - {if data doesn't fit model, change the data, not the model?}
 
===Why Rasch is better than IRT===

==== Source ====

Hobart, J., & Cano, S. (2009). Improving the evaluation of therapeutic interventions in multiple sclerosis: The role of new psychometric methods. Health Technology Assessment, 13(12). https://doi.org/10.3310/hta13120

==== Meta ====

  - created: 2021-03-11 with chapter 1
  - 2021-03-18 added chapter 2