vi conference survey sampling in social and economic research

THE KAROL ADAMIECKI UNIVERSITY OF ECONOMICS IN KATOWICE DEPARTMENT OF STATISTICS

UNIVERSITY OF ŁÓDŹ FACULTY OF ECONOMICS AND SOCIOLOGY DEPARTMENT OF STATISTICAL METHODS

POLISH STATISTICAL ASSOCIATION

VI CONFERENCE SURVEY SAMPLING IN SOCIAL AND ECONOMIC RESEARCH Katowice 21-22 September 2009

PROGRAMME AND ABSTRACTS

SPSS POLSKA – CONFERENCE SPONSOR

Katowice 2009

The 6th Conference Survey Sampling in Economic and Social Research Programme and abstracts

Monday September 21 room 335A 8.15-9.15

Registration

9.15-9.30

Opening

9.30-10.00

Invited lecture: N.T.Longford.

10.00-10.30

Chair: M.Ghosh

10.30-11.00

Coffee break

11.00-11.30

J.Wesołowski

11.30-12.00

D.Krapavickaite

12.00-12.30

A.Plikusas

Chair: J.Kordos

12.30-13.00 13.00-13.30

Coffee break Invited lecture: P.Lahiri

13.30-14.00

Chair: N.T.Longford

14.00-14.30

Lunch

14.30-15.00 15.00-15.30

J.Wywiał

15.30-16.00

J.Kordos

16.00-16.30

E.Dziwok

16.30-17.00

E.Getka-Wilczyńska

Chair:J.Wesołowski

17.00-17.30 17.30-18.00 18.00-18.30 18.30-19.00 19.00-19.30

Conference dinner

19.30-??.??

3

Tuesday September 22 room 335A 8.15-9.15

Registration

9.15-9.30 9.30-10.00

Invited lecture: M.Ghosh

10.00-10.30

Chair: P.Lahiri

10.30-11.00

Coffee break

11.00-11.30

V.Valentova

11.30-12.00

E.Pahkinen, M.Keto

12.00-12.30

T.Żądło

12.30-13.00

Chair: D.Krapavickaite Coffee break

13.00-13.30

A.Zięba, J.Kordos

13.30-14.00

B.M.Martelli

Chair: E. Pahkinen

14.00-14.30

Lunch

14.30-15.00 15.00-15.30

A.Jędrzejczak, J.Kubacki

15.30-16.00

E.Soszyńska

16.00-16.30

R.Gawlik

16.30-17.00

Conference closing

Chair: C.Domański

4

A house price index based on the potential outcomes framework NICHOLAS T. LONGFORD, SNTL and Pompeu Fabra University, Barcelona, Spain.

A house price index (HPI) is a summary of the trends in house prices. It compares the price of a typical property at one time point with the price of a similar property at another time point, such as a year ago. HPI has some features of a consumer price index (CPI). The principal difference is that a residential property is a complex product which includes not only the structure (a building or its part) and its surroundings (garden, paving, fencing and the like), but also its environment, (access to the property, proximity of shops, schools, and other services, absence of dilapidation, criminal activity and (industrial) pollution, and the like. In contrast, the items in the basket of goods for a CPI are standardised, with their qualities and functions subject to no alteration. A typical HPI is based on a register of transactions of residential properties(houses, flats, etc.), with the attributes of the properties and circumstances of the transaction (market or private sale, mortgage or paid in cash, type of tenure, etc.). The ideal HPI compares like-with-like transactions in one period with the transactions in another. This is impossible to arrange, because we can exercise no controll over the attributes of the properties involved in the transactions. This problem is addressed by the established methods by regression adjustment (hedonic house price indices) and repeatedsales analysis. The presentation will discuss the deficiencies of these, and will propose the potential outcomes framework (POF) as an approach to constructing HPI to a higher standard of integrity. In POF, the time of the transaction is regarded as the treatment, and the treatment effect is the difference in the prices paid for an identical property at one time point and another. However, a property at a fixed address is not the same consumer item as it was, say, five years ago, because of the wear and tear, maintenance and other factors (including fashion) that bring about changes, including the attributes of the environment (a new park, school, employment opportunities, reputation for petty crime, and the like). The strength of POF is that it can respond to this concern, at least in principle. A reference stock of properties is considered together with the effects of time for each property. The HPI is defined as the average treatment effect for the stock. This is practical to define on the multiplicative (log) scale, so that the calculus of percentages is simplified. An example of a proposal of HPI for New Zealand will be discussed, together with the various options of how POF can be implemented, and their connection with the methodology for missing data, and multiple imputation in particular.

5

Benchmarked Small Area Estimators WILLIAM R. BELL(a), GAURI S. DATTA(b) AND MALAY GHOSH(c) (a) US Census Bureau (b) University of Georgia (c) University of Florida,

The paper considers benchmarking issues in the context of small area estimation. Both external and internal benchmarking are considered, and some indications are provided to find the relation between the two. Optimal estimators within the class of benchmarked linear estimators are found both under external and internal benchmarking. In addition, necessary and sufficient conditions for self-benchmarking are found for an augmented model. Most results of this paper are found using the idea of orthogonal projection.

Criticalities in Applying the Neyman’s Optimality in Business Surveys: a Comparison of Selected Allocation Methods PAOLA MADDALENA CHIODINI (a)(d), RITA LIMA (c) , GIANCARLO MANZI (b)(d), BIANCA MARIA MARTELLI (c), FLAVIO VERRECCHIA (d) (a) Department of Statistics, University of Milano-Bicocca, Milan, Italy (b) Department of Economics, Business and Statistics, Universita degli Studi di Milano, Milan, Italy (c) ISAE, Rome, Italy (d) ESeC, Assago (MI), Italy.

In finite population sampling, when dealing with negligible sampling fractions (budget constraints) or when data quality is not satisfactory (e.g. frame lists, auxiliary information, etc.), a rethinking of Tchuprow’s (1924) optimality concept (later extended by Neyman, 1934) in stratified sampling is needed. In fact, the need of stratum representativeness from one side and the optimum allocation from the other are often in conflict. Furthermore, the choice of a sampling design which is rigorously respectful of the statistical theory is a difficult task in practice, especially when knowledge barriers and operational constraints are present. Sometimes a purposive design is the only possibility. This work aims at finding an optimal allocation method from a population of enterprises. This task is carried out through a simulation approach which compares different methodologies. We consider, among others, and together with the popular optimal allocation and its multivariate extension (Bethel’s algorithm -1989), the ‘Optimum Robust Allocation with Uniform Stratum Threshold – AORSU’ (Chiodini P.M., Manzi G., Verrecchia F.; 2008) and the ‘dynamic allocation’ (Buisson, 2009). These methods are suitable both for domain analyses and for the improvement of the estimates, obtained through a proxy of the stratum variability. In general, AORSU is useful when an ex-ante allocation is possible. On the other hand the dynamic allocation (i.e. ex-post) is becoming more and more of interest especially when no information about stratum variability is available. The method relies on an in itinere 6

adjustment process (i.e. an allocation based upon an estimate of the sampling variance of the first sampling data from the strata). For the latter method, even though a number of practical solutions for the survey list are available (see, e.g., Chiodini P.M., Facchinetti, Manzi G., Nai Ruscone M., Verrecchia F, 2008), the substitution of sampling units to be interviewed, which is routinely performed, is the real practical drawback, especially when organizational problems, related to target shifting, arise. The solution to this is generally not trivial. Simulations will be carried out on different databases (ASIA - ISTAT; Italian Register of enterprises - Chamber of Commerce of Milan, Italy).

REFERENCES

Bethel J. (1989) “Sample Allocation in Multivariate Survey”, Survey Methodology, Vol.15, No. 1, pp.p47-57, June. Buisson B. (2009) Estimateur et calcul de precision en cours d’enquete, priorites de relance, in Resumes des interventions, JMS2009, INSEE, 55-56. Chiodini P.M., Facchinetti S., Manzi G., Nai Ruscone M., Verrecchia F. (2008) Metodi e applicazioni per l’inferenza da popolazione finita: imprese e campionamento stratificato a selezione ordinata, ESeC. [Working paper: ESeC_WP004P_V20080714]. Chiodini P.M., Manzi G., Verrecchia F. (2008) Allocazione ottimale robusta con soglia uniforme di strato ESeC. [Working paper: ESeC_WP005P_V20080912]. Cochran W.G. (1977) Sampling Techniques, John Wiley & Sons, New York. Kish L. (1965) Survey Sampling, Wiley Classics Library, New York. Malgarini M., Margani P., Martelli B.M. (2005) New design of the ISAE Manufacturing Survey, Journal of Business Cycle Measurements and Analysis, 1, 125-142, OECD, Paris Murthy M.N. (1967) Sampling Theory and Methods, Statistical Publishing Society, Calcutta, 376-378. Rao P.S.R.S. (2000) Sampling Methodologies with Applications, Chapman & Hall, USA. Särndal, C.E.; Swensson, B.; Wretman, J. (2003) Model Assisted Survey Sampling, Springer Smith P., Pont M., Jones T. (2003) Developments in business survey methodology in the Office for National Statistics, The Statistician, 52, 257-295. Verrecchia F., Chiodini P.M., Coin D., Facchinetti S., Nai Ruscone M. (2008) Bayesian Approach for Nonresponse, in: SSBS08 - Satellite RSS 2008 conference, Southampton, UK (26-29 August 2008). [Online]. Available at http://www.s3ri.soton.ac.uk/ssbs08/programme.php]

7

Yield curve estimation: a comparison of methods with an example of Polish data EWA DZIWOK Department of Applied Mathematics, Karol Adamiecki University of Economics in Katowice,

A yield curve, understood as a plot of financial instruments’ yields against their maturities, has started to play an important role in financial markets. Rate modelling is a process of building a continuous function from the market data (securities and interest rate derivatives) based on a functional relationship between discount, spot and forward rates. The construction uses generally two types of models: parametric models (evaluated by Nelson-Siegel and Svensson) and ones based on B-splines. Both types of models give a lot of possibilities for further analysis especially for monetary policy end forecasting. The first aim of the article is to compare the methods of estimation depending on the source of data (least squares method based on rates and prices will be taken into account). An inefficiency of Polish financial market and lack of data suggests being very careful to make a model choice. This is why the second purpose is to find the best type of model for fitting the yield curve from the different assets.

Practical Problems in the Implementation of the Method of Sample Survey in the Field of Social Statistic Research RYSZARD GAWLIK Statistical Office in Cracow. Kazimierza Wyki 3, Cracow, Poland.

In my speech I would like to discuss about practical problems in the implementation of the sampling methods. I will base on two different surveys conducted by the Statistical Office: - labour force survey (LFS), which includes private households - the demand for labour, which includes national economy units. I will describe the concept of each survey, subjective scope, objective scope and sampling methods. Basing on these examples I will present problem with completeness of surveys and I will show factors which may affect on it. Also I will talk over the most frequent reasons of refusal for participation in surveys.

8

Stochastic properties of the Internet sample ELŻBIETA GETKA-WILCZYŃSKA Warsaw School of Economics, Institute of Econometrics, Division of Mathematical Statistics Al. Niepodległosci 162, 02-554 Warsaw, Poland.

Collecting data through the Internet surveys is useful either for marketing and other private research societies either statistical agencies and it has several advantages, such as low costs of collecting information, the speed of the data transmission and a possibility to monitor it. The usage of the electronic questionnaire in the Internet survey makes the interview more efficient, lowers the workload of the respondents and controls the responds’ quality. But the basic problem in the surveys over the Internet is concerned with collecting data sets according to classical methods of the sampling theory. In the Internet surveys drawing the sample is not possible and respondents are not randomly selected to the sample, but they participate in the survey with a subjective decision. The methods of the sampling theory can not be used for the data from such the samples because the probability inclusions are not known. In this paper we present basic characteristics of the Internet sample defined as the stochastic process and an example of an estimate of the rate of exponential distribution for three designs. In the first the time of the survey conducted is specified in advance and the size of the Internet sample is random value, in the second the size of the Internet sample is specified in advance and the time of the survey conducting is random value, in the third, both, the time of the survey conducting and the size of the Internet sample are specified in advance. Model based approach is proposed to estimate of parameters of the population. REFERENCES Barlow, R.E. & Proschan, F. (1965, 1996). Mathematical theory of reliability. J. Wiley and Sons, Inc. New York or Philadelphia SIAM, ISBN 0-89971-369-2 Biffignandi, S. & Pratesi, M. (2000). Modelling firm response and contact probabilities in Web surveys, Proceedings of the “Second International Conference on establishment surveys”, USA, June 2000, Buffalo Bracha, Cz. (1996). Theoretical foundation of survey sampling (in Polish), Polish Scientific Publisher, ISBN 83-01-12123-8, Warsaw Couper, M.P. & Miller, P.V. (2008). Web survey methods, Public Opinion Quarterly, Vol. 72, No. 5, 2008, Special Issue, pp. 831-835, ISSN 0033-362X Galesic, M. (2006). Dropout on the Web: effects of interest and burden experiences during an online survey, Journal of Official Statistics, Vol. 22, No. 2, June 2006, pp. 313-328, ISSN 0282-423X Getka-Wilczyńska, E. (2003). Using WWW for data collection in a national survey on scholarship granted to young resarches, Statistics in Transition, Journal of the Polish Statistical Association, Vol. 6, No. 1, June 2003, pp. 33-55, ISSN 234-7655 Getka-Wilczyńska, E. (2005). Stochastic properties of Internet data collection process, Proceedings of the 55th Session of the International Statistical Institute, CD-ROM, paper No. 998, ISBN 1-87704028-2, Australia, April 2005, ISI, Sydney Getka-Wilczyńska, E. (2007). Markov methods in test of the population lifetime, Proceedings of the 56th Session InternatTiona Statistical Institute, CD-ROM, paper No. 1708, ISBN 978-972-8859-718, Portugal, July 2007, Centro de Estatística e Aplicacŏes (CEAUL), Instituto Nacional de Estatística (INE) & ISI, Lisbon Getka-Wilczyńska, E. (2008). The Internet Sample, Statistics in Transition- new series, Journal of the Polish Statistical Association, Vol. 8, No. 3, pp. 553-560 Kingman, J.F.C. (2002): Poisson’ processes (in Polish), Polish Scientific Publishers, ISBN 83-0113534-4, Warsaw

9

Kopociński, B. (1973). Outline of the renewal and reliability theory, (in Polish), Polish Scientific Publishers, Warsaw Särndal, C.E.; Swensson, B. & Wretman, J. (1992). Model assisted Survey Sampling, Springer- Verlag New York Inc., ISBN 0-387-97528-4, New York Sołowiew, A.D. (1983). Analytical methods in reliability theory (in Polish), Technical – Scientific Publisher, ISBN 83-204-0473-8, Warsaw Tillě, Y. (2006). Sampling algorithms, Springer Science-Business Media, Inc., ISBN 10:0-387-308148; ISBN 13:978-0387-30814-2, New York Vehovar, V. (2007). Workshop on Internet Survey Methodology, Ullehammer, September 2007, http://vv.ris.org www.WebSM.org)

Estimation of Gini Coefficient for Regions from Polish Household Budget Survey Using Small Area Estimation Methods ALINA JĘDRZEJCZAK(a), JAN KUBACKI(b) (a) Chair of Statistical Methods, University of Lodz and Statistical Office in Łódź (b) Statistical Office in Łódź

In the article authors present some empirical results for the Gini coefficient estimation on the basis of the Polish Household Budget Survey obtained for regions. The direct estimation was done using Ineq package for R-project environment. The precision of direct estimation was calculated using the bootstrap technique. The small area models for the Gini coefficient was also presented. To obtain the model-based estimates of the Gini ratio for regions the EB and EBLUP techniques were applied. For EBLUP estimation SAE package for R-project environment was used

On sample allocation for effective EBLUP estimation of small area totals MAUNO KETO(a), ERKKI PAHKINEN(b) (a) Mikkeli University of Applied Sciences (b) University of Jyväskylä

The demand of regional or small area statistics produced from large-scale surveys such as Finnish unemployment survey has arisen needs for developing the tools of optimal sample allocation on the area level. For producing such tools, we use two preliminary information sources 1) complete register of auxiliary variables and 2) pre-determined small-area model (nested error regression model). From this auxiliary information we use an explanatory variable that is moderately correlated with the outcome variable the latter meaning number of unemployed persons, and we estimate the totals of this variable on NUTS2 level (provinces). Different allocation criteria as values of average absolute relative bias and efficiency are experimented by simulation studies with Finnish unemployment data. 10

Comparisons of Some Data Quality Issues in Statistical Publications in Poland in the Last Decade JAN KORDOS Formerly Warsaw School of Economics.

According to Eurostat (2003, 2007), production of high quality statistics depends on the assessment of data quality. Without a systematic assessment of data quality, the statistical office will risk losing control of the various statistical processes such as data collection, editing or weighting. Doing without data quality assessment would result in assuming that the processes cannot be further improved and that problems will always be detected without systematic analysis. At the same time, data quality assessment is a precondition for informing the users about the possible uses of the data, or which results could be published with or without a warning. Indeed, without good approach to data quality assessment statistical institutes are working in the dark and can make no justified claim to being professional and to deliver quality in the first place. The author focuses on Poland’ practice in data quality assessment in the last decade. He refers to his article published in 2001 (Kordos, 2001), where he assessed the presentation of information on sources of errors in several dissemination media: short-format publications, main report from the surveys, analytic publications, and the Internet in 1990s. He attempts to assess what happened during the last decade in data quality assessment taking into account the main sample surveys carried out by the Central Statistical Office of Poland (GUS) in 2000s. The following sample surveys are considered: Household Budget Survey, Labour Force Survey, Business Tendency Survey (industry, construction, trade), Small-Sized Enterprise Survey, Farm Structure Survey, Land Use Survey, Livestock Sample Survey, Agriculture Production Tendency Survey, Income and Living Conditions Survey (PL-SILC), Consumer Tendency Survey, Health Status Survey, Post-Enumeration Survey of Population Census. Some improvement in data quality assessment are considered.. Next, Eurostat efforts to implement data quality assessment methods are discussed. Assessing data quality is therefore one of the core aspects of a statistical institute’s work. Consequently, the European Statistics Code of Practice (Hahn and Lindén, 2006; Eurostat 2006), highlights the importance of data quality assessment. Its principles require an assessment of the various product quality components such as relevance, accuracy (sampling and non-sampling errors), timeliness and punctuality, accessibility and clarity as well as comparability and coherence. At the end, some concluding remarks are given. REFERENCES Biemer, P. and Lyberg, L. (2003): Introduction to Survey Quality. Hoboken, N.J.: Wiley. Eurostat (2003): Handbook “How to make a Quality Report”. Methodological Documents, Working Group “Assessment of quality in statistics”, Luxembourg, 2-3 October 2003. Eurostat (2007), Handbook on Data Quality Assessment Methods and Tools. Hahn, M. and Lindén, H. (2006): The European Statistics Code of Practice for a High Quality Kordos, J. (2001), Some Data Quality Issues in Statistical Publications in Poland, , Statistics in Transition, vol. 5, Nr 3, pp. 475-489. Lyberg, L. et al. (2001): Summary Report from the Leadership Group (LEG) on Quality, Proceedings of the International Conference on Quality in Official Statistics, Stockholm, Sweden, 14-15 May 2001.

11

Estimation of a total of a study variable having many zero values DANUTE KRAPAVICKAITE Institute of Mathematics and Informatics, Akademijos str. 4 , Vilnius, LT-08663, Lithuania.

The total of the values of a study variable defined for the elements of a finite population is an often used parameter of interest in survey sampling, and design-based estimation method is the most popular approach in practice. Unfortunately, it fails to obtain accurate estimates in the case of a non-negative study variable having many zero values. This kind of a variable is, for example, an area under some kind of crop that is not grown up very often in the farm, the investment of an enterprise for environmental protection that can be highly positive or equal to zero, number of hours worked by a married woman and so on. The censored regression model-based estimator of a finite population total, namely, tobit model-based estimator, has been proposed earlier. It is biased if the model assumptions are not satisfied. The presentation will be devoted to the application of the semiparametric model to a variable with many zero values, estimation of the population total by model-based and model-assisted estimators, and comparison of them with other known estimators by simulation.

Calibrated estimators under different distance measures ALEKSANDRAS PLIKUSAS Vilnius University, Naugarduko 24 , Vilnius , Lt-03225 , Lithuania.

Calibrated estimators of the finite population total are considered. Calibrated estimators use the auxiliary information in order to get more accurate estimates of the parameters. Different distance measures can be used to construct calibrated estimators. It is known, that in some cases calibrated estimators coincide with the ratio or regression estimators, which are more accurate provided the study variable is well correlated with the known auxiliary variable. Some estimators of totals as well as their approximate variances are presented in the paper. The experimental comparison of the considered estimators is made.

12

Modelling the influence of human capital on economic growth-the role of samples, influential observations and outliers ELŻBIETA SOSZYŃSKA Uniwersytet Warszawski, Centrum Badań, Polityki Naukowej i Szkolnictwa Wyższego Krakowskie Przedmieście 26/28, 00-927, Warszawa, Poland.

Several influential papers have indicated that the cross-country regression analysis shows that level or increases in educational attainment on growth output is weak or even negative. One reason is that simple cross-country regressions do not detect an effect of human capital. The effect could be hidden by a small number of unrepresentative countries, perhaps ones in which human capital accumulation has had little or no effect.Another reason is that simple application of OLS method is sometimes an inappropriate way to estimate cross-country growth regressions, especially when the results are not accompanied by a careful exploration of sample sensitivity. It may be dangerous to draw generalisations from growth regressions without investigating the likely extent of parameter heterogeneity.

Treating Statistical Data Set with Missing Values VLADIMIRA HOVORKOVA VALENTOVA Technical University of Liberec, Studentska 2, Liberec 1, 46117, Czech Republic.

Examples of statistical data treatment presented on lessons of statistics usually suppose that obtained data are randomly distributed and no values are missed. But the reality is different. Every person who works with data sets usually has to solve the problem of missing values. There are many reasons why we miss some data. The problem can be on responder’s side as well as on researcher’s side. At any rate the researcher has to decide what to do if some data miss. This paper focuses on possibilities what to do in such situations. Firstly, it is necessary to decide whether to replace the missing values or not. If we decide to replace them, we have to choose a suitable method of data imputation. We have a wide spectrum of these methods. Obviously, the used method has to fit the type of a variable. The choice depends on many factors like e.g. data variability, type of a variable, sample size, using of data etc. There is presented a simulation of research which contains all types of variables “ nominal, ordinal, interval and ratio “ in this paper. And all the variables have some missing values. We try to answer the question if it is better to leave the data as they are or to replace missing values with reference to the factors mentioned above. Afterwards the results of the both approaches are compared then. In the end we summarize advantages and disadvantages of used methods.

13

A Simulation Study of Gibbs Sampler for a Hierarchical Bayesian Model in Small Area Statistics JACEK WESOŁOWSKI Główny Urząd Statystyczny and Politechnika Warszawska, Warszawa, Poland.

We consider one of the two basic models in small area statistics - the one in which responses of units are modeled by a hierarchical system of conditional distributions of the normal and gamma type. The Bayes estimators, more precisely, approximate values of these estimators are obtained as respective posterior means via a Gibbs sampler algorithm. We lead a wide simulation analysis of the quality of such approximations. Our aim is at to study how the performance of the Gibbs sampler is influenced by such issues as: diversity of small areas, value of the model parameters, starting points of the algorithm. It appears that these aspects are also important for quality of estimators a posteriori themselves. It should be stressed that analytical results, which are known in the literature, give only very imprecise bounds on quality of Gibbs sampler approximations. While it is a common belief that the algorithm may work well in the context of small areas, no serious numerical analysis is available in the literature. The lecture, which is based on a joint work with W. Niemiro (UMK, Toruń), aims at filling this gap.

14

Simulation analysis of accuracy estimation of population mean on the basis of strategy dependent on sampling design dependent on difference of order statistic of an auxiliary variable JANUSZ WYWIAŁ Department of Statistics, University of Economics in Katowice, Bogucicka 14, 40-226, Katowice, Poland.

The paper deals with an analysis of the accuracy of the strategies for estimating the mean value of a variable under study in a fixed and finite population. A positive valued auxiliary variable is taken into account. Sampling designs dependent on quantiles are taken into account. It is proportional to the difference of two quantiles of an auxiliary variable. Next the regression type estimator with slope coefficient estimated by means of a function of order statistics is constructed. Moreover, Horvitz-Thompson, the ordinary regression strategy as well as Singh and Srivastava’s strategies are considered. The comparison of the strategies’ accuracy has been based on a computer simulation.

REFERENCES Horvitz D. G., Thompson D. J. (1952). A generalization of sampling without replacement from finite universe. Journal of the American Statistical Association, vol. 47, s. 663-685. Särndal C. E., B. Swensson, J. Wretman: (1992): Model Assisted Survey Sampling. Springer Verlag, New York-Berlin-Heidelberg- London-Paris-Tokyo-Hong Kong- Barcelona-Budapest. Singh P., Srivastava A.K. (1980): Sampling schemes providing unbiased regression estimators. Biometrika, vol. 67, 1, pp. 205-9. Wywiał J. L. (2008). Sampling design proportional to order statistic of auxiliary variable. Statistical Papers vol. 49, Nr. 2/April, pp. 277-289. Wywiał J. L. (2009). Performing quantiles in regression sampling strategy. Model Assisted Statistics and Applications vol. 4, no. 2, pp. 131-142.

15

Comparing three methods of standard error estimation for poverty measures AGNIESZKA ZIĘBA(a), JAN KORDOS(b) (a) Warsaw School of Economics, Warsaw School of Economics, Al. Niepodległosci 162, 02-554 Warszawa, Poland.

(b) Formerly Warsaw School of Economics.

Indicators of poverty and social exclusion are an essential tool for monitoring progress in the reduction of these problems. For this reason, a set of poverty measures commonly named as leaken indicators at the European Council Summit in December 2001 was established. Some of these indicators are calculated according to the Eurostat recommendations, using data from European Statistics on Income and Living Conditions (EU-SILC). Complex sample design of this survey requires approximate methods of standard error estimation, generally Taylor linearization or replication techniques. In our study three such methods for five laeken indicators are presented. We compare results of bootstrap, jackknife and linearization methods.

On some pseudo-EBLUP in the case of modeling longitudinal profiles TOMASZ ŻĄDŁO Department of Statistics, University of Economics in Katowice, Bogucicka 14, 40-226, Katowice, Poland.

Small area estimation for longitudinal data is considered. Special case of the General Linear Mixed Model with subject specific random components is assumed what is called modeling longitudinal profiles. In the case of model misspecification the Best Linear Unbiased Predictors and Empirical Best Linear Unbiased Predictors may be biased. Using robust for model misspecification estimators or predictors can solve this problem. In this class of predictors some model-assisted, design-consistent predictors called pseudo-EBLUPs are considered in the literature for data from one time period. In the paper some pseudo-EBLUP will be considered for longitudinal data.

16