Model based approach for estimating and forecasting crop statistics: Update, consolidation and improvement of AGROMET model “AGROMET Project” Working Group.

Slides:



Advertisements
Similar presentations
Introduction to parameter optimization
Advertisements

Autocorrelation Functions and ARIMA Modelling
CountrySTAT Team-I November 2014, ECO Secretariat,Teheran.
Forecasting Using the Simple Linear Regression Model and Correlation
Part II – TIME SERIES ANALYSIS C5 ARIMA (Box-Jenkins) Models
6-1 Introduction To Empirical Models 6-1 Introduction To Empirical Models.
Correlation and regression
BA 555 Practical Business Analysis
Data Sources The most sophisticated forecasting model will fail if it is applied to unreliable data Data should be reliable and accurate Data should be.
1 Spreadsheet Modeling & Decision Analysis: A Practical Introduction to Management Science, 3e by Cliff Ragsdale.
Prediction and model selection
Forecasting & Time Series Minggu 6. Learning Objectives Understand the three categories of forecasting techniques available. Become aware of the four.
Chapter 11 Multiple Regression.
Lecture 16 – Thurs, Oct. 30 Inference for Regression (Sections ): –Hypothesis Tests and Confidence Intervals for Intercept and Slope –Confidence.
© 2000 Prentice-Hall, Inc. Chap Forecasting Using the Simple Linear Regression Model and Correlation.
1 BA 555 Practical Business Analysis Review of Statistics Confidence Interval Estimation Hypothesis Testing Linear Regression Analysis Introduction Case.
Week 14 Chapter 16 – Partial Correlation and Multiple Regression and Correlation.
Slides 13b: Time-Series Models; Measuring Forecast Error
CHAPTER 18 Models for Time Series and Forecasting
BOX JENKINS METHODOLOGY
Traffic modeling and Prediction ----Linear Models
Chapter 15 Forecasting Copyright © 2011 Pearson Addison-Wesley. All rights reserved. Slides by Niels-Hugo Blunch Washington and Lee University.
Chapter 11 Simple Regression
Introduction to Forecasting COB 291 Spring Forecasting 4 A forecast is an estimate of future demand 4 Forecasts contain error 4 Forecasts can be.
Linear Trend Lines Y t = b 0 + b 1 X t Where Y t is the dependent variable being forecasted X t is the independent variable being used to explain Y. In.
© Jorge Miguel Bravo 1 Eurostat/UNECE Work Session on Demographic Projections Lee-Carter Mortality Projection with "Limit Life Table" Jorge Miguel Bravo.
Chapter 8: Regression Analysis PowerPoint Slides Prepared By: Alan Olinsky Bryant University Management Science: The Art of Modeling with Spreadsheets,
Q2010, Helsinki Development and implementation of quality and performance indicators for frame creation and imputation Kornélia Mag László Kajdi Q2010,
Inferences in Regression and Correlation Analysis Ayona Chatterjee Spring 2008 Math 4803/5803.
ESTIMATING & FORECASTING DEMAND Chapter 4 slide 1 Regression Analysis estimates the equation that best fits the data and measures whether the relationship.
1 Spreadsheet Modeling & Decision Analysis: A Practical Introduction to Management Science, 3e by Cliff Ragsdale.
Forecasting supply chain requirements
1 Calculation of unit value indices at Eurostat Training course on Trade Indices Beirut, December 2009 European Commission, DG Eurostat Unit G3 International.
Time Series Analysis and Forecasting
6. Evaluation of measuring tools: validity Psychometrics. 2012/13. Group A (English)
2.4 Units of Measurement and Functional Form -Two important econometric issues are: 1) Changing measurement -When does scaling variables have an effect.
Linear Trend Lines = b 0 + b 1 X t Where is the dependent variable being forecasted X t is the independent variable being used to explain Y. In Linear.
© Copyright McGraw-Hill Correlation and Regression CHAPTER 10.
Introduction to Time Series Analysis
Eurostat – Unit D5 Key indicators for European policies Third International Seminar on Early Warning and Business Cycle Indicators Annotated outline of.
Copyright © 2011 by The McGraw-Hill Companies, Inc. All rights reserved. McGraw-Hill/Irwin Model Building and Model Diagnostics Chapter 15.
Welcome to MM305 Unit 5 Seminar Prof Greg Forecasting.
R. Ty Jones Director of Institutional Research Columbia Basin College PNAIRP Annual Conference Portland, Oregon November 7, 2012 R. Ty Jones Director of.
K. Ensor, STAT Spring 2005 Estimation of AR models Assume for now mean is 0. Estimate parameters of the model, including the noise variace. –Least.
Forecasting Parameters of a firm (input, output and products)
The Box-Jenkins (ARIMA) Methodology
Irwin/McGraw-Hill © Andrew F. Siegel, 1997 and l Chapter 14 l Time Series: Understanding Changes over Time.
Chapter 14 Introduction to Regression Analysis. Objectives Regression Analysis Uses of Regression Analysis Method of Least Squares Difference between.
7-8-March 2011 Task Force "Organic farming statistics"-Luxembourg, 7-8 March Item 4 Harmonised questionnaire for data collection: State of the art.
1 General Recommendations of the DIME Task Force on Accuracy WG on HBS, Luxembourg, 13 May 2011.
Managerial Decision Modeling 6 th edition Cliff T. Ragsdale.
1 Ka-fu Wong University of Hong Kong A Brief Review of Probability, Statistics, and Regression for Forecasting.
Economics 173 Business Statistics Lecture 28 © Fall 2001, Professor J. Petry
Ljubljana, | Slide 1 Groundwater Quality Assessment Determination of chemical status and assessment on individual sites Austrian experience.
G2 Crop CIS meeting Ispra, May 14 – 15, 2012 Presented by: Institute of Geodesy and Cartography.
Labour Cost Index (LCI) Calculation of the LCI in Denmark.
Forecasting. Model with indicator variables The choice of a forecasting technique depends on the components identified in the time series. The techniques.
Lecture 9 Forecasting. Introduction to Forecasting * * * * * * * * o o o o o o o o Model 1Model 2 Which model performs better? There are many forecasting.
Stats Methods at IC Lecture 3: Regression.
Inference about the slope parameter and correlation
Multiple Imputation using SOLAS for Missing Data Analysis
Financial Econometrics Lecture Notes 2
Evaluation of measuring tools: validity
Week 14 Chapter 16 – Partial Correlation and Multiple Regression and Correlation.
Imputation in UNECE Statistical Databases: Principles and Practices
Correlation and Regression
L. Isella, A. Karvounaraki (JRC) D. Karlis (AUEB)
BA 275 Quantitative Business Methods
3.2. SIMPLE LINEAR REGRESSION
Implementation of the Bayesian approach to imputation at SORS Zvone Klun and Rudi Seljak Statistical Office of the Republic of Slovenia Oslo, September.
Presentation transcript:

Model based approach for estimating and forecasting crop statistics: Update, consolidation and improvement of AGROMET model “AGROMET Project” Working Group Meeting on Crop Statistics October 25th, Luxembourg

DevStat Agenda 1. Introduction: Project Goals and current AGROMET model. 2. Solutions: a new model. A.Key features and Improvements B.Imputation of missing data C.Model selection D.Model validation E.Set-Aside effect of A 3. Results and examples. 4. Conclusions and comments.

DevStat Project goals 1. Introduction To provide forecasts for crops, vegetables and fruit production on every country member at the EU, based on historical data on both observables provided to Eurostat periodically: harvest production (H) and arable area (A). To generate a SAS application for the statistical analysis, and use the SAS Enterprise Guide to create its interface. To generate documentation on the application and form users.

DevStat Project Implementation 1. Introduction DEVSTAT and UMH-CIO are the members of the consortium providing services to Eurostat under the Framework contract number S Within this framework contract, the consortium received the request for services ESTAT E0/24 for the updating, consolidation and improvement of current AGROMET model, according to the goals stated in the previous slide. This presentation shows the main features of the project and the current state of the development and implementation of the new model.

DevStat Current Agromet Model: main features Eurostat has been using the AGROMET MODEL, programmed within FAME, to provide forecasts for areas, yield and production on crops, on twelve of the countries at the EU, and also on its aggregate. Yield is calculated by dividing the observable vaiables Production and Area. Estimates of Yield and Area are produced based on the last 10 years, by using: Linear regression Quadratic regression ARIMA(1,1,1) Production estimate is obtained by multiplying Yield and Area estimates. 1. Introduction

DevStat Current Agromet Model: limitations The current AGROMET model: Constraints forecasts from fixed estimation models for each product, whatever the country and prediction year. Constraints some forecasts to predictions based on lines or parables, also assuming independency of the historical available data series (inconsistent assumption on time correlated data series) Excludes forecasting in all these cases (combinations product-country) with 2 or more missing values in the data series from the last 10 years. Focuses on forecasting the non-observable variable YIELD to provide forecasts on the observable HARVEST PRODUCTION. Does not provide forecasting error measures. Only covers crops on 12 from the 27 current UE members. 1. Introduction

DevStat Key features of the new model 2. Solutions HARVEST PRODUCTION (H) and AREA (A) are the relevant observable variables in the analysis and model fitting. Prediction data: 10 years time series. IMPROVEMENTS An Imputation procedure for missing data values has been developed. Trend models reasonably substitute fixed Agromet models. Relationship between the observables A and H is used for H prediction whenever possible. Inclusion of an automatic model selection criteria for different fitted models. Error measures are provided on the forecasts. Set-aside specifications on A can be used to predict the effect on H.

DevStat Imputation procedure 1) Regression/Inverse regression of H on A. 2) Moving Average Estimate of lag 1 2. Solutions

DevStat Imputation procedure 2. Solutions HARVEST AREA 1) Regression/Inverse regression of H on A.

DevStat Imputation procedure 2. Solutions 2) Moving Average Estimate of lag 1

DevStat Imputation results 2. Solutions

DevStat Model especification Case 1 Case 2 Case 3 4 or more NA data Average of last 3 years 3 or less NA data ARIMA (p,d,q) no NA data on H and A A: ARIMA (p,d,q) Regression models for H on A with ARMA (p,q) errors 2. Solutions

DevStat Model selection Automatic selection rule for the best model among the ARIMA(p,d,q) and ARMA(p,q) models. Flexible and optimal adaptation to data, whatever the product, country and behaviour along time. Intrinsic validation mechanism: the best model provides minimum deviation between observed and estimated. These facts are distinctive improvements with respect to the previous AGROMET model. 2. Solutions

DevStat Validation The validation index VAL is defined as the “percentage of cases whose confidence or credible interval does get the observed value at the prediction year”. This validation index is assessed at two levels: COUNTRY-TYPE-PREDICTION YEAR (country VAL index) PRODUCT-PREDICTION YEAR (product VAL index) 2. Solutions

DevStat Set-aside on Area 2. Solutions In order to produce forecasts of HARVEST under land restrictions, two possibilities are considered: When a regression model is available, to predict HARVEST based in the regression model, with AREA, as explanatory variable. If not, to predict HARVEST on a proportional basis, by applying the same percentage of reduction for arable land.

DevStat Interface in SAS E-G 3. Results Selection of Country

DevStat Interface in SAS E-G 3. Results Selection of Type of product

DevStat Interface in SAS E-G 3. Results Selection of product

DevStat Interface in SAS E-G 3. Results Selection of Prediction year

DevStat Interface in SAS E-G 3. Results Selection of Variable to predict

DevStat Estimation results: example case 1 Case 1 4 or more NA data on A and H Average of last 3 years 3. Results

DevStat Estimation results: example case 2 Case 2 3 or less NA data on A and H H: ARIMA (p,d,q) A: ARIMA (p,d,q) 3. Results

DevStat Estimation results: example case 3 Case 3 no NA data on A and H ARIMA(p,d,q) on A Regression models for H on A with ARMA (p,q) errors 3. Results

DevStat Validation Results: product VAL index 3. Results

DevStat Validation Results: country VAL index 3. Results

DevStat Conclusions 4. Conclusions The AGROMET model has been improved: It uses more appropriate time series models It adapts the available data, whatever the country, product or prediction year, always with the optimal model selected. It provides forecasting errors. Its interface allows for selection on type of product, product, country, prediction year, set-aside. It accommodates a set-aside restriction on A It has been programmed under SAS Enterprise Guide.

DevStat Comments 4. Conclusions THANKS FOR YOUR ATTENTION