StatXact10PROCs_online

< Contents * Index >>>

Preface

Welcome to StatXact PROCs, a statistical package for Exact Nonparametric Inference , also known as Permutational Inference . StatXact 10 PROCs is a special version of the StatXact 10 software package, developed for SAS users. Previously, this package used to be referred as Proc-StatXact. The goal of StatXact PROCs is to enable statisticians and data analysts to make reliable inferences by exact and Monte Carlo methods when their data are sparse, heavily tied, or skewed, and the accuracy of the corresponding large sample theory is in doubt. Very efficient numerical algorithms are used in StatXact PROCs. For small data sets, this ensures quick computation of exact p-values, confidence intervals, and power. If a data set is too large for the exact algorithms, StatXact PROCs computes Monte Carlo estimates of the exact p-values to any desired level of accuracy. If a data set is too large for both exact and Monte Carlo inference, it is almost certainly large enough for the asymptotic theory to work accurately. In effect, StatXact PROCs provides a comprehensive set of nonparametric procedures that perform reliably for small, intermediate and large data sets. Aligned Mid-Rank Score Test (Stratified) . Aligned Mid-rank test is now available for two independent samples with stratified data. An option for this test is added in PROC TWOSAMPL. The details of this test are available in section 10.6 . Lepage Test . The Lepage test is available for detecting the shift as well as dispersion alternatives. This test is a combination of the Wilcoxon-Mann-Whitney test and the Ansari-Bradley test which are meant for shift and dispersion alternatives respectively. This test has been added in PROC TWOSAMPL. The details of this test are available in section 10.15 . Boschloo’s Test . The Boschloo’s unconditional test is available for comparing proportions of two independent binomials. This test has been added in PROC BINOMIAL. The details of this test are available in subsection 17.3.4 . Also, The Power and Sample Size calculations for Boschloo’s test are explained in subsection 37.3.3 An option for the Power and SS computation is added in PROC SXPOWERBIN. Tests for Correlated RxC Contingency Tables . The correlated tests are available for the unordered, singly ordered and doubly ordered RxC tables. The details of these tests are available in Chapter 29 . Various correlated tests are added in PROC CORELATED. MID-P Corrected Confidence Interval . Mid-p corrected confidence intervals are displayed for various tests under ’One Sample rates and proportions’ (Chapter 14 ), ’C ordered Binomials’ (Chapter 20 ), ’Two ordered Multinomials’ (Chapter 21 ) and ’trend tests for clustered data’ (Chapter 28 ) Family Based Association Test . This new test is available in StatXact 10 procs for INFERENCE FOR GENETIC DATA. This test is appropriate for family based data when the investigative marker has two or more alleles. In StatXact10, we allow up to four alleles. A new proc namely PROC FBAT is added in this version to run this test. The details of this test are available in chapter 34 . Depending upon the version of SAS, there could be a conflict of StatXact PROCs Proc names with SAS PROC names. For example, PROC ASSOCIAT is in StatXact PROCs as well as in SAS/Enterprise Miner. PROC POWERBIN and PROC POWERMLT in StatXact PROCs conflict with PROC POWER in SAS/STAT. The new features in StatXact 10 PROCs are the following:

New in StatXact 10 PROCs

Procedure Naming Conven- tion in StatXact PROCs

In order to circumvent such errors, we are providing the following facility:

i

Preface

Made with