Sunday, November 13, 2016

Regression Modeling in Practice - About the Data



Sample - The sample of 17543 craters is from A NEW MARTIAN GLOBAL CRATER DATABASE TO 1.0 KM. by S. J. Robbins and B. M. Hynek (http://about.sjrdesign.net/thesis.html) which provides a database for Mars that contains 378,540 individual craters statistically complete for diameters greater than or equal to 1 km. This database includes location,diameter,depth and ejecta morphology .My sample will be limited to those craters with a described ejecta morphology and a diameter less than 50 km.


Data Collection Procedure – Observational data was collected from satellites in orbit around Mars over a period of years and identified manually. Satellite data included THEMIS Day IR mosaics, Viking MDIM 2.1 where gaps exist in THEMIS coverage, CTX mosaics in isolated cases, and MOLA data.


Measures -  Size and position were determined using ARC GIS software. Morphology was determined using ARC GIS and IGOR software. Latitude and Longitude were binned into Martian quadrangles as defined by the United State Geological Survey(Astrogeology Science Center). Ejecta types have been binned into groups of first layer ejecta morphology types.

Sunday, October 30, 2016

Data Analysis Tools - Moderation

  This week I wanted to find out if the linear relationship between depth and diameter was true for all Ejecta types. Essentially, is Ejecta type(and the underlying structure that produces it) a moderator of depth and diameter? In this case I binned the types to single, double and multiple layer. (SLE, DLE, MLE)
  The charts below show that depth and diameter maintain the positive linear relationship regardless of the Ejecta types chosen. The DLE and MLE types are strongest at .78 and .79 respectively, with the SLE being somewhat less at .61.

SAS code is at the end.