Journal of the Royal Statistical Society: Series A (Statistics in Society)

Cover image for Vol. 180 Issue 2

Edited By: J. Carpenter and H. Goldstein

Impact Factor: 1.702

ISI Journal Citation Reports © Ranking: 2015: 13/49 (Social Sciences Mathematical Methods); 24/123 (Statistics & Probability)

Online ISSN: 1467-985X

Associated Title(s): Journal of the Royal Statistical Society: Series B (Statistical Methodology), Journal of the Royal Statistical Society: Series C (Applied Statistics), Significance

175:3


Log-optimal economic evaluation of probability forecasts, by D. J. Johnstone, Journal of the Royal Statistical Society, Series A, Statistics in Society, Volume 175, (2012), pages 661–689

The file Johnstone(LogOptimal#1) does the calculations that appear in the simple numerical illustration in Tables 1 and 2, and these are fully detailed in the paper.

The file Johnstone(LogOptimal#2) contains the football data, and the file Johnstone(LogOptimal#3) contains the weather data.

D. J. Johnstone
Sydney University
Business School
University of Sydney
Sydney NSW 2006
Australia

E-mail: david.johnstone@sydney.edu.au

Dataset

Assessing gross domestic product and inflation probability forecasts derived from the Bank of England fan charts, by J. W. Galbraith and S. van Norden, Journal of the Royal Statistical Society, Series A, Statistics in Society, Volume 175, (2012), part 3, pages 713–727.

Quick Start (Windows)

  1. Unless you are a university researcher doing academic research, you must stop and get written permission from the authors of these programs prior to using them.
  2. Extract all files into a new subdirectory.
  3. Start GAUSS and change the working directory to the one containing the files.
  4. Run Infl PIT.gau and GDP PIT.gau. This requires the GAUSSPlot application module. If you do not have GAUSSPlot, you may comment out the last part of the program, starting with the line library gaussplot;
  5. Run kernels.gau and kernels_GDP.gau.

Raw Data

The underlying data were downloaded from the Bank of England (BoE) website (http://www.bankofengland.co.uk/publications/inflationreport/irprobab.htm), which provides the three parameters (roughly speaking, the mean, variance and skew) determining the probability distribution for every forecast at every forecast horizon. The names and format of the files below reflect those that were in use by the BoE in 2010Q1; at present they should be easy for users to update.

The files are:

cpiinternet.xls Parameters for MPC CPI Inflation Projections from February 2004
gdpinternet.xls Parameters for MPC GDP Growth Projections August 1997 - August 2007
gdpbankforecast.xls Parameters for MPC GDP Growth Projections based on Bank Estimates of Past Growth from August 2007
historicalforecastdata.xls Parameters for Bank of England RPIX Inflation Projections February 1993 - May 1997
rpixinternet.xls Parameters for MPC RPIX Inflation Projections August 1997 - February 2004

Data for GDP, CPI and RPIX were taken from the O¢ ce of National Statistics (ONS) website www.statistics.gov.uk.

UKRPIX.xls Dataset Name: rpi1q - Retail Prices Index: quarterly index numbers of retail prices 1948-2007 (RPI) (RPIX) Series: RPIX - All items excluding mortgage interest payments - Index: January 13 1987=100

UKCPI.xls UK CPI Annual Percentage Change

UKGDP.xls Dataset Name: natpc2 - National accounts: GDP: expenditure chained volume measures at market prices, Seasonally adjusted; £ million at chained volume measures. Series name: ABMI - Gross domestic product at market prices

Programs

These data files are then transformed into probabilistic forecasts and out comes by two GAUSS programs. Both programs make the use of some customized GAUSS procedures, which are described after the programs. Note that some of these procedures and programs are copyrighted by Ergodic Quantitative Consulting Inc. They may be freely used by university researchers in academic research provided that they cite this paper. Use in governments or corporations is strictly prohibited without prior written agreement from Ergodic Quantitative Consulting Inc. The file newey-west.g is due to Leonardo Bartolini and Charles Kramer of the IMF; it is included here for convenience and users should consult that file for details and limitations on its use.

GDP PIT.gau This program reads data from gdpinternet.xls, gdpbankfore cast.xls and UKGDP.xls, calculates PIT transforms of the outcomes, and creates histograms of their distributions. The histogram is shown in the right column of Figure 1 and requires AUSSPlot plus some minor interactive reformatting to adjust the proportions and colour flood settings. The program also produces six output files in Excel format.

  • ?GDP25 gives the probabilistic forecast that 4Q real GDP growth 2.5%.
  • GDPPrIT? gives the forecast cdf of the reported 4Q real GDP growth.
  • The prefix in the above file names indicates the forecast used: M for market interest rates, F for fixed interest rates.
  • The suffix in the above file names indicates the measure of GDP growth used: P for first-release and V for fixed-vintage (2010Q1 in this case.)

In.PIT.gau This program reads data from historicalforecastdata.xls, rpixin ternet.xls, cpiinternet.xls, UKCPI.xls, and UKRPIX.xls, and creates histograms of their distributions. The histogram is shown in the right column of Figure 1 and requires GAUSSPlot plus some minor interactive refor matting to adjust the proportions and colour .ood settings. The program also produces six output files in Excel format, where ? = M for market interest rates, F for fixed interest rates.

  • ?PiPrIT gives the forecast cdf of the reported 4Q Inflation.
  • ?PiTarget gives the probabilistic forecast that 4Q Inflation Target.
  • ?BPiTarget gives the probabilistic forecast that 4Q Inflation will be within the Target Bands.

kernels.gau This program reads the output files produced by In.Pit.gau and produces tables of statistical analysis and many graphs. The results given in the paper in Table 1 are those for Market-based forecasts. The graphs are those for Figures 1 (left and center columns) through 3.

kernels_GDP.gau This program reads the output files produced by GDP Pit.gau and produces tables of statistical analysis and many graphs. The results given in the paper in Table 1 are those for Market-based forecasts. The graphs are those for Figures 1 (left and center columns) through 3.

Procedure Library

BoE2010.lcg This GAUSS library file allows the above GAUSS programs to find the relevant procedures. It references the various *.g files found in the zip file. They are

  • binorm.g - –cdfbinorm() – - FanChartCDF() – - FanReport()
  • boeparse.g – - BoEParse() - –xstats()
  • alibrationgraph.g - –CalibrationGraph()
  • datagraf.g - datagraph()
  • jwgsvn.g – - locmxnw() – - locmxll() –- locmxcv() –- locmxcom() –- TrimNWK() –- NormalKernel() –- kernspecboot() k
  • kernel.g - calkernel()
  • kernspecsimonma.g - –NormalKernel() - –kernspecsimonma() - –gzsma()
  • loadboe.g - LoadBoE()
  • newey-west.g (due to Leonardo Bartolini and Charles Kramer of the IMF) - newey_west()
  • plotecdf.g –- ecdf() - –redoecdfplot() – - plotecdf() –- ResGraph()

Dataset (215KB)

Improved probabilistic prediction of healthcare performance indicators using bidirectional smoothing models, by H. E. Jones and D. J. Spiegelhalter, Journal of the Royal Statistical Society, Series A, Statistics in Society, Volume 175, (2012), part 3, pages 729 – 747.

WinBUGS model code for fitting the Poisson hierarchical AR(1) model

The model code in file supp-mat is based on that of Lin et al. (2009), but has been extended to include random walks for the population parameters mu_t and log(tau_t), allowing each of these to change flexibly over time and predictions to be formulated automatically.

Due to concerns about some of the 'uninformative' prior distributions used by Lin et al. (2009) we also changed these to alternatives which we believe are more appropriate. Further, we include code for one and two step ahead predictions and corresponding predictive p-values.

The model code requires the following data inputs:
* T = number of time periods
* m = number of healthcare providers
* R = m x T matrix of observed counts
* E = m x T matrix of expected counts

Hayley E. Jones
School of Social and Community Medicine
University of Bristol
Canynge Hall
39 Whatley Road
Bristol BS8 2PS
UK

E-mail: hayley.jones@bristol.ac.uk

Dataset (2KB)

How real is mobility between low pay, high pay and non-employment?, by D. Pavlopoulos, R. Muffels and J. K. Vermunt, Journal of the Royal Statistical Society, Series A, Statistics in Society, Volume 175, (2012), part 3, pages 749 – 773.

Description of the datasets:

Britisch Household Panel Survey, waves 1-14 Access obtainable through the UK Data Archive, see http://www.esds.ac.uk/findingData/bhps.asp

German Socio-Economic Panel, waves 8-21 Access obtainable through the DIW Berlin, see http://www.diw.de/en/diw_02.c.222829.en/access.html

Dutch Socio-Economic Panel, waves 7-18. Access obtainable through Statistics Netherlands, see http://www.cbs.nl/nl-NL/menu/methoden/dataverzameling/sociaal-economisch-panelonderzoek-sep.htm

The analysis of the data was done in LatentGOLD, see http://statisticalinnovations.com/products/latentgold.html

More information can be obtained by the principal author:
Dimitris Pavlopoulos
VU University Amsterdam
Faculty of Social Sciences dept. of Sociology
De Boelelaan 1081
(visiting address Metropolitan Building, Buitenveldertselaan 3)
1081 HV Amsterdam
The Netherlands

E-mail: D.Pavlopoulos@vu.nl

Datasets (.zip 2KB)

SEARCH

SEARCH BY CITATION