2016-03-15 - stages before secondary data analysis

[andkov]

2016-03-15

Notes during meeting:

• keep domain map local to each workshop
• participants MUST bring their data in WIDE format

  Definition for wide format

Variable names in the wide data format the last (four) characters must be:

• "_" - underscore
• "t" - lower-caset "t" for time point
• "01" - number of the wave as double digit
• The name may contain more than one underscore.
• case does not matter
• only letters,numbers, and underscores are allowed

Participant must provide for two things:

I. Initiation stage

• A wide format data (see definition for wide format) that passes data-validator (script Pass data validator with )
• requires a personal sit-down meeting (virtual meeting) with a IALSA representative to run the data-validator script agains the data of the longitudinal study. Alternatively, the participant may choose to validate the dataset himself by following instruction on applying a data-validation scripts.
• built-it in the IalsaSynthesis package
• optionally, this function produces a report log that serves as verification document for passing the data validator. The produced report acknowledges the variables and the features of the data set
• this meeting is the chance to install the necessary software on the participants, RStudio, Github, R packages.
• the laptop/machine be used for this verification step is recommended to be brought to the workshop.

  II. Pre-Conference stage

• only one person (email) will be allowed to input into the Pre-Conference per study.
• survey invitation can be resend
• survey can be edited and completed in steps
• specify for each time-variant variable the time points for it should be modelled.

  III. Catalog

( each row in the catalog = a model from a study) This stage is governed by a script. It follows automatically, once PCS is "lock-in" for a study.

Catalog will be sequentially populated using the information provided in the Pre-Conference Survey and the custom templates. This process will be accomplished in four phases.

• Phase 1. Using the info from PCS, R scripts populate the values in the variables that define model specification (e.g. model_type, subgroup, process A, process B, ect)
• Phase 2. Using the data from Phase 1 and a prototypical Mplus script, R script creates .inp file to be used by Mplus for model estimation.
• Phase 3. Using .inp files create in Phase 2, estimation is conducted. .out file are produced.
• Phase 4. The .out files created in Phase 3 are parsed and recorded as individual variables.

• transition from phase 2 to phase 3 occurs on participants' computer that has access to the source data ( possibly included into IaslaSynthesis)

  Homework for Maleeha :

• combine into and variables into a single REDCap form

  Homework for Andrey:

• example of data set ( in .csv format) that will be used for modelling.
• confirm that it's ok to use MAP data set (subset) for our examples.

[andkov]

Later in the day, sessioned with @wibeasley to develop the parser

Notes during the meeting

To do list:

• change timeN to tN in the prototype ( for brevity)
• work with a inconsistent wave pattern (e.g. 1, 2, 4, 5 to anticipate irregularity issues in creation algorithms)
• make the collector function take in a single model output
• add commented tokens that spell model specification (e.g. yaml header)