Consolidate GSCAM and GSCA into I-GSCA
Opened this issue · 0 comments
emstruong commented
- Ensure that I-GSCA implementation reproduces current implementation of GSCA and GSCA-M
- Archive old GSCA and GSCA-M code
- Replace GSCA and GSCA-M with I-GSCA
- Check if you can re-produce the loadings and path coefficients from ALS, by standardizing the weights, then following the rest of the cSEM algorithm to get loadings and path-coefficients.
- What do "revised for simulation" mean in igsca_sim.m? What is it supposed to be?
- See if we can reproduce matlab results if we standardize weights -> get loadings + path-coefficients via cSEM's normal way.
- Maybe adjust GSCA_m to also be another
.approach_weights
option and to also require the.disattenuate
to beTRUE
to run? - Fix issue regarding models with common factors and composites
- Complete
.GSCA_modes
functionality and documentation -
X
andW$Data
are not the same in00_foreman.R
for IGSCA... Why? Does this matter? Can we simplify it? - Ensure that
calculateFIT()
,calculateFIT_m()
andcalculateFIT_s()
are all compatible with GSCA_m and GSCA - Add
calculateAFIT()
- Make IGSCA exit early to side-step unnecessary computation using an
if (){}
statement, but still output a standardforeman()
return to maintain compatibility - Fix
calculateReliability()
compatibility with IGSCA, GSCA_m and more, if necessary. - Allow random initial weights in IGSCA. Should it be passed to initializeIgscaEstimates? Or should initializeIgscaEstimates be bypassed?
- Add tests for IGSCA compatibility when a construct only has one indicator (see GSCA_m implementation in calculate_weights.R)
- Allow the estimation of both canonical and nomological composites in I-GSCA
- Appropriate bypasses when only 1 construct of any kind is modelled, no structural model, etc