Dear all,
On Wednesday 27th February the RSS Leeds/Bradford local group will be hosting a meeting on
"Statistics in Education".
Further details can be found on our webpage:
http://www.maths.leeds.ac.uk/statistics/rss/current.html#EDUCATION
All welcome!
Regards, Paul
===================================================================
Dr. Paul D. Baxter
Secretary/Treasurer, RSS Leeds/Bradford Local Group,
Department of Statistics, University of Leeds, Leeds, LS2 9JT, UK.
-------------------------------------------------------------------
Leeds/Bradford: Wednesday 27 February, 2pm, Leeds University
Harvey Goldstein (University of Bristol).
The effect of pupil mobility on school differences in educational achievement.
The recently introduced National Pupil Database in England allows the tracking of every child
through the compulsory phases of the state education system. The data from Key Stage 2 for three
Local Education Authorities are studied, following cohorts of pupils through their schooling. The
mobility of pupils among schools is studied in detail using multiple membership multilevel models
that include prior achievement and other predictors and the results are compared with traditional
'value added' approaches that ignore pupil mobility. The analysis also includes a cross
classification of junior and infant schools attended. The results suggest that some existing
conclusions about schooling effects may need to be revised.
Robert West (University of Leeds).
Moderation of student assignment marks using MLM.
For large classes of students, prompt marking of essay assignments cannot be achieved without using
multiple markers. Even with strict marking guidelines, some markers can be more generous than
others, or more likely to detect student errors, or more tolerant of minor weaknesses. Consequently
moderation is required. An objective methodology is sought to achieve this.
It is good practice to employ either second marking or double marking of a proportion of student
scripts. Here double marking will be considered. Student scripts are submitted with only student
numbers to identify them so that markers assess scripts anonymously. Around 10% of scripts are
selected randomly and duplicated. These are assigned so that markers can be assessed against each
other. In addition, further double/triple marking is undertaken for some scripts that lie close to
the pass/fail or key grade boundaries. A multilevel model is employed with scripts in the upper
level. Markers are nested within scripts and cross-over is permitted.
The methodology is demonstrated with a student assessment where there are 263 students and 14
markers. A total of 45 scripts are double marked and 18 are triple marked. The assignment was a new
one for the course and the marking scheme was new also. Some markers were more experienced than
others. The variance due to markers was seen to be relatively large compared to the variance due to
different student abilities. Fitted (random) marker effects were used to provide the necessary
adjustment to final student mark allocation.
The relatively high marker variation shows the need for moderation in order to achieve an adjustment
to permit a fair mark to be assigned to each student. The cross-over design induced by double and
triple marking substantiates the marker effects. This design could be extended to include a random
selection of results from another student cohort should moderation be required against that cohort
as well as between markers. The results of the moderation were fed back to markers so that they
could see their performance compared with their peers. This may lead to closer marker agreement for
future assignments.
The meeting will be held at Leeds University’s Roger Stevens Building in RSLT 19, starting at 2pm.
Refreshments will be available from 1.30pm in the foyer of level 9 of the School of Mathematics.
|