매월 조사되는 주택 가격 변동률의 이상치 탐색 방법에 관한 연구 = Outlier Detection Methods for Monthly Rate of Housing Price

다국어 초록 (Multilingual Abstract)

In statistical theory, an outlier is a value that is numerically distant from overall pattern of a distribution. It may be a meaningful observation, but it comes from an error in survey, data entry and process in most cases. Detection and handling are needed because outliers by errors debase the statistical quality leading to biased parameter estimation. Generally, traditional Box-plot or Z-score are very useful for univariate outlier detection and a Median rule could be applied in the non-Gaussian case. These methods calculate the tolerance interval that defines the range of acceptable observation values. Outlier detection for periodic surveys would consider the past view, because it is based on a ratio of value comparing the current time with previous time. If time period, however, is short, a state to get many unchanged values can occur. In this case, the ratio is centered at 1, and therefore outlier detection method reflecting this factor is required. This paper considers Quartile Method with power transformation and Hidiroglou-Berthelot(1986) method that is efficient in periodic data. The methods were applied to housing sales price. We suggest an outlier detection method for real-world data. In addition, we also analyzed data using Tukey Algorithm of United Kingdom``s office of National Statistic(ONS).

참고문헌 (Reference)

1 Hoaglin, D. C., "Understanding robust and exploratory data analysis" Wiley 1983

2 Thompson, K. J., "Statistical methods for developing ratio edit tolerances for economic data" 15 : 517-535, 1999

3 Hidiroglou, M. A., "Statistical editing and imputation for periodic business surveys" 12 : 73-84, 1986

4 Carling, K., "Resistant outlier rules and the non-Gaussian case" 33 : 249-258, 2000

5 Lee, H., "Outliers in Sample Surveys" Statistics Canada 1992

6 Liu, R. Y., "On a notion of data depth based on random simplices" 18 : 405-414, 1990

7 Granquist, L., "Improving the traditional editing process, In Business Survey Methods" John Wiley & Sons 1995

8 John W. Tukey, "Exploratory Data Analysis" Addison-Wesley 1977

9 Hubert, M., "An adjusted boxplot for skewed distributions" 52 : 5186-5201, 2008

10 Barnerjee, S., "A simple univariate outlier identification procedure designed for large samples" 36 : 249-263, 2007

11 Liu, R. Y., "A quality index based on data depth and multivariate rank tests" 88 : 252-260, 1993

12 Serfling, R., "A depth function and a scale curve based on spatial quantiles, In Statistical data analysis based on the LI-Norm and related methods" Birkhauser 25-38, 2002

13 Brys, G., "A comparison of some new measures of skewness, In Developments in Roubust Statistics (ICORS 2001)" Springer-Verlag 98-113, 2003