haku: @author Mcleay, S. / yhteensä: 10
viite: 2 / 10
| Tekijä: | McLeay, S. Omar, A. |
| Otsikko: | The sensitivity of prediction models to the non-normality of bounded and unbounded financial ratios |
| Lehti: | British Accounting Review
2000 : JUN, VOL. 32:2, p. 213-230 |
| Asiasana: | Forecasting Models |
| Vapaa asiasana: | Financial ratios |
| Kieli: | eng |
| Tiivistelmä: | This paper distinguishes between 2 types of financial ratio: (i) unbounded ratios which may take either negative or positive values and where extreme values are possible in both tails of the distribution, and (ii) bounded ratios which may take positive values only and which are characterized by extreme values in the right-hand tail only. The empirical study that is reported in the paper indicates how such transformations based on ratio form can lead to an improvement in the classification accuracy of prediction models. A comparison with other techniques that have been proposed for accommodating outliers in the raw data shows that the transformations proposed in this paper, which retain the information in the data set as a whole, result in greater accuracy than the ex-post data deletion procedures recommended elsewhere. |
SCIMA