Lat value added knowledge share part 2
This presentation is the property of its rightful owner.
Sponsored Links
1 / 28

LAT Value Added Knowledge Share Part 2 PowerPoint PPT Presentation


  • 66 Views
  • Uploaded on
  • Presentation posted in: General

LAT Value Added Knowledge Share Part 2. YPLA Strategic Analysis & Research team. Championing Young People’s Learning. The Methodology. Championing Young People’s Learning. Value Added Specification. It’s long and it’s got some complex equations

Download Presentation

LAT Value Added Knowledge Share Part 2

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Lat value added knowledge share part 2

LAT Value AddedKnowledge SharePart 2

YPLA Strategic Analysis & Research team

Championing Young People’s Learning


The methodology

The Methodology

Championing Young People’s Learning


Value added specification

Value Added Specification

It’s long and it’s got some complex equations

Producing a version with a bit more plain English in it

Methodology independently reviewed by NfER

Underlying concepts are complex…

…but individual steps are not so bad

Championing Young People’s Learning


Multi level modelling more detail

Multi Level Modelling: More detail

MLM applied at qualification & subject level (e.g. A level History)

Attempting to fit a polynomial equation

Complexity of equation based on number of records & insts

Where there are very few records things are grouped together at SSA level

Championing Young People’s Learning


Multi level modelling polynomials

Multi Level Modelling: Polynomials

Quadratic (order 2)

A + Bx + Cx2

80 to 500 cases

500 to 5000 cases

5000+ cases

Cubic (order 3)

A + Bx + Cx2 + Dx3

Quartic (order 4)

A + Bx + Cx2 + Dx3 + Ex4

Championing Young People’s Learning


Multi level modelling polynomials1

Multi Level Modelling: Polynomials

Quadratic (order 2)

A + Bx + Cx2

80 to 500 cases

500 to 5000 cases

5000+ cases

Cubic (order 3)

A + Bx + Cx2 + Dx3

Quartic (order 4)

A + Bx + Cx2 + Dx3 + Ex4

A, B, C, D and E are known as ‘gamma’ coefficients

Championing Young People’s Learning


Multi level modelling provider lines

Multi Level Modelling: Provider Lines

Championing Young People’s Learning


Multi level modelling provider lines1

Multi Level Modelling: Provider Lines

  • National Line= 10 – 5x + 0.5x2 + 0.01x3

  • Provider A Line= 12 + 10x + 0.5x2 + 0.01x3

  • Provider B Line=1000 – 3x + 0.5x2 + 0.01x3

  • Provider C Line= -400 + 10x + 0.5x2 + 0.01x3

Championing Young People’s Learning


Multi level modelling equations

Multi Level Modelling: Equations

National Line

Provider Line Variation

Distribution Of Results

(Variance)

Error

Championing Young People’s Learning


Multi level modelling s and r

Multi Level Modelling (S+ and R)

Cannot be done with SQL or version of SPSS we have

R & S+ are programmes that support MLM calculation

Data Service use S+ but has licence issues

YPLA getting open source programme R onto estate

Both use the same programming language

Code used works in a similar way to SPSS

Championing Young People’s Learning


Multi level modelling basic r syntax

Multi Level Modelling Basic R Syntax

base<-read.table('C:\\DriveD\\LAT VA\\Q_111_S12330.dat',

header = TRUE, fill = TRUE)

library (lme4)

a<-lmer(POINTS~PRIORC+PRIOR2C+PRIOR3C+PRIOR4C

+(PRIORC+PRIOR2C|LAESTAB),data=base)

fixef(a)

Championing Young People’s Learning


The calculation process

The Calculation Process

Championing Young People’s Learning


Step 1 data from fft

Step 1: Data from FFT

Received in SPSS format

FFT can advise on issues with the data

Includes data fields on:

Provider code (UPIN & LAESTAB)

Qualification type (A09 and LAT VA qual codes)

Prior attainment and outcome attainment

Learner Names

Championing Young People’s Learning


Step 2 check and sort data

Step 2: Check and sort data

Data service do a range of checks on the data

Drop any qualifications that are too small to include

Check all providers we expect to be included are in the data

Produce “centred variables”

They then split into lots of small files (one per qual & subject)

Championing Young People’s Learning


Centred variables

Centred Variables

Many equations include quartic term (i.e. x to the power 4)

For a prior attainment score of 58 this is a big number

58 x 58 x 58 x 58 = 11,316,496

To make the MLM calculation run quicker the variables are centred

This means smaller numbers are used

Involves some fiddly calculations but just basic maths

If you hear reference to “Prior C”, “catalyst file” or “beta variables” these are interim steps used in this centring process

Championing Young People’s Learning


Step 3 apply mlm

Step 3: Apply MLM

Data is fed into S+ programme one file after another

This will fit the national line and give details on distribution of results

Gives an output as text file

These text files are then grouped together using a compiler routine

Championing Young People’s Learning


Step 4 check solutions

Step 4: Check solutions

The individual solutions are fed into a spreadsheet to check whether they look reasonable

This spreadsheet is known as the “Batch LAT”

Original FfE version was very complex (Over 6000 lines of code)

For 2009/10 we will be using a simplified version (1000 lines of code)

General checking by eye

Some mathematical checks too (positive definite matrix)

Produces “decentered coefficients”

Championing Young People’s Learning


Step 5 re apply mlm

Step 5: Re-apply MLM

If the solution does work then apply a lower order equation

…or group data up to SSA level

Then re-check solution

End point of this is national lines for all of the qualifications

Championing Young People’s Learning


Step 6 upload data to online lat

Step 6: Upload data to Online LAT

The national lines and individual provider data are uploaded into the Online LAT

This undertakes the same calculations as the Batch LAT

It also generates reports for all providers

These are viewable through the provider gateway

Championing Young People’s Learning


Step 7 extract result files

Step 7: Extract result files

Bulk data files can be exported from the Online LAT

The “new style” reports will be generated using these files

Results files sent to OfSted to be used in their reports

Results files uploaded to SQL for use by YPLA

Results for 2007/08 and 2008/09 are on MISVS001 in LSC_MI_DB_PUB(the filenames start LATVA_)

Championing Young People’s Learning


The ready reckoner

The ReadyReckoner

Championing Young People’s Learning


Ready reckoner

Ready Reckoner

  • Providers were keen to get an early view of LAT VA scores

  • Excel spreadsheet that allows them to model their own data

  • Due for release in early October

  • Uses data from LAT VA 08/09 amended data release

  • First line support by Data Service’s Service Desk

  • Second line support by YPLA


Input data

Input Data


Output data

Output Data


Work scheduled for the next 6 months

Work scheduled for the next 6 months

Championing Young People’s Learning


Timescales

Timescales

  • Unamended run mid November

  • Updated LAT Handbook and communications at same time

  • Amended run planned for mid January


Conclusion

Conclusion

Championing Young People’s Learning


Lat value added knowledge share part 2

Remember the basics

A

B

A level result

C

D

National average achievement

E

E

D

C

B

A

Average GCSE grade


  • Login