2016-06-01 Writing Session I.

[andkov]

involved: @wibeasley , @Maleeha

Agenda

• typed verbal description of the whole and the parts of the map represented by the image below.
• this is the conceptual workflow map produced for the first CAR software testing ( #49) conducted on 2016-05-26 at UVic
• as part of the preparation for a methodological paper describing software for CAR, our current conceptualization and verbalization of it will be prepared in a poster format to be presented at Canadian Psychological Association on 2016-06-09 at Victoria, BC

[andkov]

Writing during the session

Writing session for CAR poster

• .data and .csv are language-agnostic, tabulated data files
• .inp and .out are free-form, text file
• .csv belong to both private and public.
• catalog is a datasetin which each row represents one model per study.
• columns in this datasets are populated sequentially and are grouped into four sections:
• Part I - contains the responses of the drivers to the PCS as they described their data
• Part II - contains model specification as Mplus syntax (that can be persisted in a file)
• Part III - contains model output from MPlus
• Part IV - columns with parameters parsed from Part III.

PCS is REDCap survey that interacts with the driver to obtain relevant description of the study 's dataset and characteristics.

Creator Script run on server. After the responses to PCS are entered, the Creator populates (values of the variables in) Part I of the Catalog.

PART I.

Translator. Script run on server. Using Part I it Mplus model syntax, which is uploaded to Part II.

PART II.

Transcriber. Script run on each driver's local machine. Downloads model syntax from Part II of the catalog, and saves it as an .inp file on the driver's local machine.

Cruncher. Script run on each driver's local machine. Reads the catalog and determines which models still needs to be estimated. {After verifying that the .inp file (and .dat) exists locally...} It programmatically calls a local installation of Mplus, which uses the local .dat and .inp files. It returns an .out file containing the model solution.

Horseman. Script run on each driver's local machine. Uploads the contents of the .out files to Part III.

PART III.

Parser. Script run on server. Extracts the critical characteristics of the model solution from the Mplus output (such as parameter estimates, fit indices, and the convergence status). For each model, the resulting individual scalar values are saved as separate columns in a single row of Part IV.

PART IV.

Prometheus. Script run on server. Copies the entire catalog as a CSV on the driver's local machine. This disconnected CSV allows the participants to later run reports and meta-analyses with more independence.

Assets

• R, RStudio, and various packages
• IalsaSynthesis R package developed for CAR workshops
• Mplus and MplusAutomation
• REDCap
• Git and GitHub

see a stand-along document with the above notes at https://github.com/IALSA/ialsa-2015-portland-stencil/blob/master/documentation/2016-06-01.md

Random notes during meeting

• sort the forest plot? - no, not a single thing to sort on.
• homework for Andrey: check the logic of degrees of freedom computation for the fixed effects forest plots (https://github.com/IALSA/IALSA-2015-Portland/issues/127)
• homework for Andrey: create an issue for Will with info needed to pair the colors between the outcome-space map to the forest plots.
• Remember the architecture https://github.com/IALSA/IalsaSynthesis/blob/master/libs/documents/architecture.md of the IalsaSynthesis, it has many relevant words

[andkov]

The above verbalizations will be used to populate poster text #51