L.4 SAS Programs and Output for Example Data

L.4 SAS Programs and Output for Example Data

This Section provides SAS programs to analyze the example data sets given

herein Appendix L. Each program includes all analyses from the corresponding

decision tree that would be performed using SAS. While it is certainly possible

to conduct the statistical analysis of a data set in a stepwise fashion, we find it

much more efficient to perform all analyses at once, and then select the

appropriate results based on the steps in the decision tree. Power calculations

are provided in addition to the decision tree analyses.

SAS statements in the sections that follow are given in uppercase letters

(although this is not required for SAS). Comments within the body of the

programs are in upper and lowercase letters in the following format: /* Comment

line. */. Every SAS statement must end with a semicolon, but several statements

may be included on the same line. The programs are designed for the analysis of

Appendix L example data but can be used with other data sets after minor modi-

fications. Investigators wishing to use these programs should have some

familiarity with SAS. SAS output follows each program; the output has been

edited to remove much of the nonessential information.

We recommend that data analysis reports include at least the following:

Number of replicates, mean and SE for each treatment.

Treatment of less-than detection limit data, if any.

Results of tests of assumptions.

Data transformation used, if any.

Name of statistical hypothesis testing procedure, its calculated test

statistic and associated probability, and conclusion reached regarding the

null hypothesis.

Minimum significant difference or some other indication of power for a

parametric LSD test or t-test.

L.4.1 Program WATTOX.SAS for water column toxicity test data

analysis

WATTOX.SAS is a program to compare dilution water survival vs. 100

percent elutriate survival, using an arcsine-square root transformation on the

data. The program performs all statistical analyses in Figure L-1. Included in

these analyses are: mean survival for dilution water and elutriates,

Shapiro-Wilk's Test for normality, t-test for equal or unequal variances, and a t-

test for unequal variances on data converted to rankits. Refer to the decision tree

in Figure L-1 to determine which test results should be used. Minimum

significant difference and some other power calculations for the parametric t-test

L40

Appendix L Selected Resource Documents