ECONOMICS|economic analysis|statistics
- samhail asamtóiteach Tagairt Faomhadh an téarma seo mar chuid de Thionscadal Lex
- ga
- asymptotic model
- en
- Nóta "Many standard statistical tests are based on large sample or asymptotic theory. If a design calls forfewer than 5 experimental units per concentration, such large sample statistical methods may not beappropriate. In addition, if the measurement is sufficiently crude, then a large proportion of the measuredresponses have the same value, or are very restricted in the range of values, so that tests based on apresumed continuous distribution may not be accurate. In these situations, an exact permutation-basedmethodology may be appropriate. While universally appropriate criteria are difficult to formulate, a simplerule that should flag most cases of concern is to use exact methods when any of the following conditionsexists: (1) at least 30% of the responses have the same value; (2) at least 50% of the responses have one oftwo values; (3) at least 65% of the responses have one of three values. StatXact and SAS are readilyavailable software packages that provide exact versions of many useful tests, such as the Jonckheere-Terpstra and Mann-Whitney tests.OECD. Environment directorate. OECD Series on Testing and Assessment. Number 54. CURRENT APPROACHES IN THE STATISTICAL ANALYSIS OF ECOTOXICITY DATA: AGUIDANCE TO APPLICATION - ANNEXES http://www.olis.oecd.org/olis/2006doc.nsf/LinkTo/NT00000C3E/$FILE/JT03208635.PDF [14.7.2010]"