Page 1553 - SUBSEC October 2017_Neat

APPENDIX II

Description of Statistical Procedures

The procedure for generating assessed grades must first be governed by a policy as agreed
to by either a specialist or general committee. In this current circumstance, CXC has
followed two principles as established in 2010.

Having been advised of such principles, the statistical procedures were as follow:

1) Review each case by candidate to make sure they meet the criteria to be given an
assessed score. Of the hardship cases reported due to weather, 234 met the criterion
for an assessed grade based on the established principles.
2) Generate a population grade and score distribution with respect to the territory
(Jamaica) with the hardship cases.
3) Complete cross-correlation analysis to make sure that there exists a relationship
between and in some cases among the scores being predicted.
4) Determine the distribution functions of each subject paper at the candidate level.
5) Once the distribution function has been determined, conduct regression analysis
using both linear and non-linear techniques.
6) Generate a best fit equation from the regression analysis and adjust the parameters
of the equation to reduce the squared residuals if necessary.
7) Validate the model by randomly selecting a sample of candidates with complete
information to predict their scores.
8) After the validation process, test for stability and reliability by running various
iterations of the same models on different examination years.
9) Complete a quantile analysis in order to validate the consistency of the Centre’s
performance.
10) Generate a percentile table to present the lower and upper bounds of the score
bands in order to assign grades based on established cut-off scores.

64