SAS Commands

Basic SAS Commands
Daren B. H. Cline
Statistics Department
Texas A&M University
26 May 1999
© (1999) Daren B. H. Cline
All rights reserved

Basic SAS Commands
i
Table of Contents
1. Introduction....................................................................................................................1
2. The Datastep ..................................................................................................................3
3. Procedures
ANOVA...................................................................................................................5
CHART....................................................................................................................6
CORR.......................................................................................................................7
FREQ .......................................................................................................................7
GLM.........................................................................................................................8
LOGISTIC ...............................................................................................................9
MEANS..................................................................................................................10
NPAR1WAY .........................................................................................................11
PLOT......................................................................................................................11
PRINT....................................................................................................................12
RANK ....................................................................................................................12
REG........................................................................................................................12
RSQUARE.............................................................................................................13
SORT .....................................................................................................................14
TTEST....................................................................................................................14
UNIVARIATE.......................................................................................................14
VARCOMP............................................................................................................15
4. Other Commands
BY..........................................................................................................................15
LABEL...................................................................................................................15
OPTION.................................................................................................................16
RUN.......................................................................................................................16
TITLE ....................................................................................................................16
5. Operators, Expressions and Functions.........................................................................16
6. Interactive Data Analysis (INSIGHT) .........................................................................17

Basic SAS Commands
1
1. Introduction
For more information, refer to the SAS Introductory Guide or the SAS/STAT Users
Guide, vol. 1 and 2 (© SAS Institute Inc., Cary NC). (See also the SAS Practical User's
Guide, © Computing Services Center.)
This guide provides description and usage for the SAS command language as it pertains
to inputting and analyzing data. Commands are generally submitted through the
PROGRAM EDITOR or with batch files. The interactive approach (INSIGHT) can
duplicate most of the basic analyses available in the command language, but certainly not
all. Thus, in practice one will use both the command language and INSIGHT. At present
the guide discusses INSIGHT only briefly (in Section 6).
GETTING STARTED: University microcomputers which are publicly available (in the
Open Access Labs) already have SAS installed. You may install SAS on a computer in
your department if your department obtains a license for each installation. Or you may
obtain free software and license to install it on your personally owned computer, as long
as you are a student. To obtain the license, go to the Software Evaluation & Loan
Library (SELL) in 113 Teague Building (862-4104).
Pay close attention to the installation instructions and to the hard disk requirements. For
most statistical analyses, including those covered here, the following modules are
adequate: CORE, BASE, STAT, INSIGHT and HELP. These modules currently require
at least 50MB of hard disk space.
After opening SAS, it initially provides three windows called PROGRAM EDITOR,
LOG and OUTPUT. Each of these may be reached at any time from the Globals menu.
Discussion of their use follow.
PROGRAM EDITOR: Program Editor Window.
Commands can be entered one or more at a time or read in from a file. Use the File menu
to open a source file (file.sas) or to save it. (The Tool Bar also has buttons for these
operations.) It is always wise to save your commands in a file for future reference.
Once a set of commands has been entered, edited and saved, they may be submitted from
the Locals menu or by clicking on the Submit button (or on F8). Make sure there is a
run statement at the end of the submitted text to ensure SAS executes everything. There
are two possibilities. If any text is highlighted then just that text is submitted and the
PROGRAM EDITOR remains unchanged. However, if no text is highlighted then all the
text is submitted and is cleared from the PROGRAM EDITOR. To retrieve text which
has been cleared (for example, to correct a mistake and resubmit it) use Recall Text from
the Locals menu (or F4).
When SAS processes the submitted text it puts messages into the LOG window and data
analysis results into the OUTPUT window.

Basic SAS Commands
2
LOG: Log Window.
This window contains messages describing the processing of the commands. The
messages include error messages, indicate whether data was input or output correctly and
provide the dataset names. It should be frequently reviewed to ensure proper analyses.
OUTPUT: Output Window.
All the analyses are provided in this window. Note that the results cannot be edited in
this window. For this reason you may wish to save it (to file.lst) and edit it in an editor of
your choice. Unwanted output may first be cleared by choosing Clear Text in the Edit
menu (or Ctrl-E).
HELP: Help Menu.
The Help menu contains all the basic information for executing SAS commands. Select
Help-SAS System Main Menu. Information on the SAS DATA Step, for example, is
found under SAS Language. All the basic procedures for analyzing data are found under
Modeling and Analysis Tools.
The rest of this section provides a few basic preliminaries about the SAS language .
PROCEDURES: SAS uses a number of procedures (identified with proc) to analyze
data. In addition, the datastep (which begin with data) is used to input and manipulate
data. The datastep and the procedures are discussed below in Sections 2 and 3,
respectively.
VARIABLE and DATASET NAMES: The names of variables and of SAS datasets must
consist of 8 or fewer alphanumeric characters; underscored blanks (_) also are allowed.
SAS does not distinguish between lower and upper case letters except in the value of a
character variable, a label or a title. In the examples below, SAS commands will be
given in lower case (e.g., proc reg). The names of variables and datasets will be in
upper case (e.g., X, DATASET1).
SEPARATING COMMANDS: Every command ends with a semicolon (;). First-time
users often find that a missing or misplaced semicolon is a cause of many errors. SAS
treats multiple spaces and blank lines as a single space so you may space commands to
suit you. Also, you may put more than one command (with semicolons) on the same line.
COMMENTS: If an asterisk (*) follows a semicolon (;) then everything between the
asterisk and the next semicolon is treated as commentary and is not interpreted by SAS.
Comments will be used to explain commands in the examples below.

Basic SAS Commands
3
2. The Datastep
DATA STATEMENT: A datastep is used to input, create, modify or combine SAS
datasets. Since most SAS procedures require a SAS dataset, this is usually the first step
other than specifying options or titles. Datasteps may also appear later in the commands.
A datastep usually contains three primary commands: data (to name the dataset and
signal SAS that a datastep begins), input (to input the values of variables from the
user's dataset) and output (to output an observation to the SAS dataset). The output
statement is often implicit and therefore optional. Between the input and output
commands, new variables can be created using ordinary mathematical expressions.
Looping can also be used. Other commands (set and merge) can be used for
modifying existing SAS datasets, but they will not be discussed here.
SAS DATASETS: SAS datasets are created using SAS. They consist of a sequence of
observations. Each observation contains the values of one or more variables named by
the user. The internal variable _N_ is the observation number, starting with _N_ = 1 for
the first observation.
USER'S DATASET: In the simplest case, the user inputs a dataset in which each line
corresponds to one observation of the SAS dataset and contains all the information
required to create that observation. More general scenarios are possible, however, as the
examples below illustrate. The user's dataset can follow the datastep or it can be in a
separate file. In the former case, the command cards must follow the datastep and the
data must begin on the line immediately after the cards command. If the data are in a
separate file, say file.dat, then an infile command is required. For example, use
data DATASET;
infile 'file.dat'; * Input data file;
INPUT STATEMENT: The input command is used to input data. It is usually easiest
to input in free format, one observation per line. However, other approaches are often
useful. See the examples below. Note that a $ must follow a character variable name.
EXPRESSIONS in the DATASTEP: Many ordinary mathematical expressions can be
used to compute new quantities. Logical expressions are indicated with parentheses (e.g.,
(X>=25)) and can be treated as ordinary numbers with values 0 (false) and 1 (true).
Also, conditional (if) statements and looping (do) statements may be used. The
examples below illustrate some uses of expressions. Section 5 lists common operators
and functions.
EXAMPLE DATASTEPS: These examples are just a small selection of the capabilities
of the datastep in SAS. Indeed, some users use SAS solely for data manipulation. The
first example is of the simplest form where each line of the user's dataset corresponds to
one observation of the SAS dataset. SAS executes the input command once for each
observation and in such a case the output command is implicit and not needed.

Basic SAS Commands
4
data DATASET1; * Create a dataset named DATASET1;
input NAME $15. SEX $ WT HT; * NAME & SEX have character values;
* NAME must have 15 characters;
cards; * The data begins on the next line;
J. Adams M 151 69
M. B. Finster F 104 63
K. C. Lodge M 208 73
M. J. Newsome M 185 69
Q. Prevoli F 132 66
;;; * Indicate end of data;
The semicolons at the end of the data are optional, but are useful both as visual markers
and to ensure SAS does not misinterpret any lines to follow.
The next example illustrates a formatted input where the columns for each variable are
identified. This is especially useful if the user's dataset contains more variables per
observation than are needed, but it requires the dataset to be strictly formatted. In this
example, the variables are input from a data file and are given labels. The if statement
is used to select only the observations where the patient is male.
infile 'patients.dat'; * Input data file;
input NAME $ 1-15 SEX $ 20 WT 30-35 HT 60-65;
if SEX='M'; * Keep only male patients;
label NAME='Name of Patient'
WT='Weight at Admission'
HT='Height at Admission'; * Label variables for plots;
Our third example shows how more than one observation can be read from the same line
of data. We also show the calculation of new variables.
input WIDTH HEIGHT LENGTH @@; * Continue reading on same line;
VOL=WIDTH*HEIGHT*LENGTH;
LOGVOL=log10(VOL);
LARGE=((VOL>20) or (LENGTH>5)); * LARGE is binary (0 or 1);
cards; * The data begins on the next line;
3.1 2.0 4.0 1.1 1.8 1.8 1.5 1.5 2.0
1.2 1.0 3.0 1.2 1.5 5.8 2.5 2.5 3.0
2.4 2.0 2.8 1.2 1.6 1.8 1.0 1.2 2.0
;;; * Indicate end of data (9 obs.);
The fourth example uses an array to input multiple observations per line which have
common values for some variables. In this case the first two variables (TEMP and
HUMIDITY) are common for five values of the third variable (STRENGTH). Missing
values are indicated with a . (period). Note that the output command is explicitly
given.
data DATASET4;
array X(I) X1-X5; * Define array X with length 5;
input TEMP HUMIDITY X1-X5;
do over X; * Begin a loop. I=1,5 is implicit;

Basic SAS Commands
5
RUN=I; STRENGTH=X; * Obtain RUN and STRENGTH;
output; * Output one observation each loop;
end; * End the loop;
drop I X1-X5; * Drop variables not needed;
cards; * 30 obs., including 2 missing;
75 70 14.3 15.9 13.8 15.7 15.2
75 80 13.6 15.3 14.8 13.2 14.6
75 90 12.3 14.7 . 15.1 13.5
85 70 14.8 15.6 14.8 15.9 16.0
85 80 14.3 13.9 15.5 15.6 15.3
85 90 13.3 . 13.4 12.7 12.5
;;;
The final example demonstrates a more explicit form of looping in order to simulate a
random sample of 50 normal data.
data DATASET5;
input MEAN STDDEV; * Input parameters;
do I=1,50; * Begin loop;
X=MEAN+STDDEV*normal(0); * normal(0) is a std. normal r.v.;
output; end; * Output X and end loop;
drop MEAN STDDEV;
cards;
40 15
;;;
3. Procedures
The most commonly used statistical procedures are given below with the required syntax
and primary options. This is meant to provide basic useful information and is not at all
comprehensive. Consult the manuals for more options and examples.
ANOVA: Analysis of Variance.
For balanced experimental designs, ANOVA is the most efficient analysis, but it can not
produce residuals (see instead GLM, or INSIGHT in Section 6). The syntax is
proc anova data=dataset; class variables;
model response variable = factors and interactions;
means effects/options;
For example, if DATASET4 shown above did not have missing data the following could
be used.
proc anova data=DATASET4;
class TEMP HUMIDITY;
model STRENGTH=TEMP|HUMIDITY; * Factorial model with interaction;
means TEMP/duncan; * Duncan's test on TEMP means;
All of the factors in the model statement must appear in the class statement. The
interactions can be given explicitly or implicitly. Examples include:

Basic SAS Commands
6
model Y=A B; * Factors A and B, no interaction;
model Y=A B A*B; * Factors A and B, and interaction A*B;
model Y=A|B; * Factors A and B, and interaction A*B;
model Y=A|B C; * Factors A, B and C, interaction A*B;
model Y=A B(C); * Factors A and B, and C within B;
The effects in the means statement can be any effect included in the model statement.
Means and standard errors are given for each main effect level and for each crossed effect
level. Multiple comparisons are performed for the main effects according to the option
requested. Options for the means statement include:
bon Bonferroni test
duncan Duncan's multiple range test
lsd Fisher's least significance difference test (default)
tukey Tukey's studentized range test
CHART: Frequency Bar Charts.
This provides a quick histogram-like chart, with more flexibility than the stem-leaf plots
of UNIVARIATE. (See also INSIGHT in Section 6.) For eight or more intervals, the
horizontal plots are usually better. The syntax is
proc chart data=dataset; hbar variables/options;
For example,
proc chart data=DATASET4;
hbar STRENGTH/type=pct midpoints=12 to 20 by 2;
Use vbar instead of hbar for a chart with vertical bars. Options to determine how the
scale of the bars is determined include
axis=value maximum value of the bar axis
type=freq bar length represents the frequency (default)
type=cfreq bar length represents the cumulative frequency
type=pct bar length represents the relative frequency
type=cpct bar length represents the cumulative relative frequency
By default SAS will choose the categories or the intervals for the chart. To specify them
instead use the option
midpoints=values
If values are numerical then the interval endpoints are chosen halfway between the
specified values. Equally spaced values result in a chart with equal length intervals
(except possibly the end intervals). The values can also be nominal values of a
categorical variable. Examples are

Basic SAS Commands
7
midpoints = 1 3 5 7 9; intervals are [min.,2), [2,4), [4,6), [8,max.)
midpoints = 1 to 9 by 2; same as above
midpoints = 1 3 7 15; intervals are [min.,2), [2,5), [5,11), [11,max.)
midpoints = A B C; categories are A, B and C
Side-by-side charts can be given for each subgroup defined by a group variable by using
the option
group=variable
CORR: Correlation Matrix.
For multivariate data, the sample correlations are provided as well as the sample means
and standard deviations of each variable. The syntax is
proc corr data=dataset options; var variables;
Options include
pearson Pearson or usual correlation (default)
spearman Spearman rank correlation
FREQ: Frequency Table.
This tabulates frequencies and/or relative frequencies according to categorical variables
given by the user. It is especially useful for cross-tabulation (when there are two or more
categorical variables) and it can do the chi-square contingency or homogeneity test. The
command is given by
proc freq data=dataset; weight frequency variable;
var variables;
The weight or frequency variable is used when the dataset consists of counts for each of
the subclasses, rather than the individual observations. To do the contingency test or the
homogeneity test use
tables class variable*class variable/options;
For example,
proc freq data=DATASET4; var TEMP HUMIDITY;
tables TEMP*HUMIDITY/expected chisq;
The options include
cellchi2 chi-square term for each cell
chisq Pearson's chi-square test
expected expected frequency for each cell under null hypothesis

Basic SAS Commands
8
GLM: General Linear Models.
This conducts the analysis of variance for general linear models, including those for
experimental designs and models with covariates. GLM should be used instead of
ANOVA when the experimental design is not balanced. GLM can also be used for
multiple regression but usually REG is the better procedure. Residuals can be obtained
for a residuals analysis. (INSIGHT can perform the basic analyses of GLM interactively,
but not the multiple comparisons or special tests.) The syntax for GLM is
proc glm data=dataset; class classification variables;
model response variable = predictor variables/options;
output out=new dataset keyword = name;
The predictor variables can be either numerical or class variables, but if any are the latter
they must be identified in the class statement. Interactions may be included as in
ANOVA (see ANOVA above). Additionally, quadratic and interactive terms with
numerical variables may be included as, e.g., X*X, A*X and X*Y.
Additional statements which may used in GLM, but which must appear after the model
statement, include the means, lsmeans, estimate and contrast statements.
These are discussed in more detail below.
We provide two examples. The first involves an experimental design with unbalanced
data (i.e., unequal sample sizes).
proc glm data=DATASET4; class TEMP HUMIDITY;
model STRENGTH=TEMP|HUMIDITY; * Factorial model with interaction;
lsmeans TEMP|HUMIDITY/stderr tdiff;
* Compare all cell means;
The second example is a model with both categorical and numerical variables. This
situation is sometimes called covariate analysis or regression with unequal slopes.
proc glm data=DATASET1; class SEX;
model WT=SEX HT SEX*HT/ss3 solution;
* The SEX*HT term allows for the slope to depend on SEX;
output out=DATANEW r=RESIDUAL p=PREDICTD;
estimate 'Avg Male Wt at 70 In' SEX HT SEX*HT 0 1 70 0 70;
* The coefficients for SEX are 0 (Female) and 1 (Male);
estimate 'Diff Avg Wt at 70 In' SEX HT SEX*HT 1 -1 70 70 -70;
* The coefficients for SEX are 1 (Female) and -1 (Male);
Options for the model statement include
i prints the inverse of the crossproducts matrix, X'X-1
noint suppresses the intercept term (default is to include it)
solution prints parameter estimates
ss1 prints the sequential sums of squares, according to the sequence of
predictors

Basic SAS Commands
9
ss3 prints the partial sums of squares for each predictor
xpx prints the crossproducts matrix, X'X
The options ss2 and ss4 are for certain exceptional situations and generally should be
avoided.
Certain values may be produced by the output statement and appended to the dataset
(thus forming a new dataset). These are indicated by keyword above and include
p values predicted by the fitted model
r residuals from the fitted model
For experimental designs the population means are estimated with the means and/or
lsmeans statements. Their syntax is
lsmeans effects/options;
means effects/options;
The options for the means statement are the same as those for ANOVA (see ANOVA).
Generally, if the experimental design is not balanced then the lsmeans statement should
be used. The options for lsmeans include
pdiff p-values for all pairwise comparisons
stderr standard errors of the estimates
tdiff t-statistics and p-values for all pairwise comparisons
Specific hypothesis tests and estimates are obtained with the contrast and estimate
statements, respectively. The syntax is
contrast 'label' effects values;
estimate 'label' effects values;
In these statements the label should be no more than 20 characters. For a class variables
(i.e., a factor) the values are a list of coefficients, one for each level of the factor, in
ascending order of the levels. For numerical variables the values are usually particular
values for those variables. The rules are somewhat complicated for handling effects
which are not included in the statement so it is often best just to include them all, with
zeros as needed. (See the second example above.)
LOGISTIC: Logistic Regression.
This procedure fits logistic regression models to a categorical response variable. Logistic
regression can be quite sophisticated, depending on the assumptions. We just discuss the
bare bones here. Note that the response is assumed to be from ordered categories. The
model fits the cumulative probabilities for those categories. If the category labels are not
in alphabetic or numeric order then the data should be grouped by category and ordered
appropriately. The syntax for LOGISTIC is as follows

Basic SAS Commands
10
proc logistic data=dataset options;
output out=new dataset keyword = name;
In the proc logistic statement options include
descending use if the categories are in reverse order (recommended if the
response is binary)
notsorted use when response categories are not in alphanumeric order and
SAS should not sort them
simple prints simple statistics for each predictor variable
Options in the model statement can include
cl provide confidence limits for parameters
corrb print correlation matrix of parameter estimates
covb print covariance matrix of parameter estimates
ctable classification table if the response is binary
noint suppress intercept in the model
risklimits provide confidence limits for odds ratios
rsquare provide R-square of model fit
Keywords allowed in the output statement include
p predicted cumulative probability for the response category
reschi residual based on Pearson's chi
resdev deviance residual
There are many other options available.
MEANS: Summary Statistics.
This procedure calculates summary statistics, such as sample mean and variance. It
differs from UNIVARIATE in that the resulting statistics may be output to a new dataset.
The syntax is
proc means data=dataset keyword list; var variables;
class classification variable;
output out=dataset keyword(s)=name(s);
The keywords in keyword list are names for various summary statistics such N, MEAN,
STD, MIN, MAX and VAR. These same keywords are used in the output statement to
assign names to the variables in the output dataset. For example,
proc means data=DATASET4 n mean std; var STRENGTH;
class TEMP HUMIDITY; out=DATANEW n=N mean=MEAN std=STD;

Basic SAS Commands
11
NPAR1WAY: Kruskal-Wallis Test.
Distribution free procedures for the one-way classification model such as the Kruskal-
Wallis test are obtained by
proc npar1way options data=dataset;
var response variable; class classification variable;
The Kruskal-Wallis test is also called the Wilcoxon test when there are two groups. To
use this test to compare weights for the two sexes in DATASET1,
proc npar1way wilcoxon data=DATASET1; var WT; class SEX;
Options include
anova the usual one-way analysis of variance (F-test)
wilcoxon the Kruskal-Wallis test
Other options may be used, depending on assumptions.
PLOT: Character Plot.
Although the resolution is not high, reasonable scatterplots can be obtained. (See also
INSIGHT in Section 6.) The syntax for PLOT is
proc plot data=dataset;
plot Y variables * X variables = plot variable/options;
The plot variable is optional. The default will be a letter indicating the number of data
plotted at that location (A for one observation, B for two observations, etc.). The plot
variable can be the value of a variable in the dataset or it can be a single character in
quotes, e.g. '*'. Options for PLOT include
overlay to overlay two or more plots
haxis = values identify values for the horizontal (X) axis
href = value identify where to draw the vertical (Y) axis
vaxis = values identify values for the vertical (Y) axis
vref = value identify where to draw the horizontal (X) axis
Some examples are
proc plot data=DATASET1;
plot WT*HT/haxis=60 to 75 by 5; * plot of WT vs. HT;
plot WT*HT='*'; * plots with character *;
plot WT*HT=SEX; * plot character is the value of SEX;
plot STRENGTH*(TEMP HUMIDITY);
* two plots with different X variables;

Basic SAS Commands
12
proc plot data=DATASET;
plot Y1*X='*' Y2*X='+'/overlay; * overlays two plots;
PRINT: Listing of Data.
This lists each observation with its values for the variables specified. NOTE: This
procedure does not send output to the printer. The syntax is
proc print data=dataset; var variables;
If no variables are provided then all the variables are output by default.
RANK: Data Ranking.
The data ranked in order of value (alphabetic order for character data). One option will
produce the corresponding quantiles of the normal distribution. SAS creates a new
dataset with all the old variables plus the rank values of each ranked variable. To rank
the data use
proc rank data=dataset out=new dataset;
var variables; ranks ranked variables;
For example
proc rank data=DATASET1 out=DATANEW;
var HT WT; ranks HT_RANK WT_RANK;
To obtain normal quantiles (called N_QUANT here) for the variable X and then to plot X
in a normal quantile plot use
proc rank data=DATASET out=DATANEW normal=blom;
var X; ranks N_QUANT;
proc plot data=DATANEW; plot X*N_QUANT;
REG: Regression Analysis.
Estimates of the coefficients of a multiple regression and their standard errors are
provided. It also provides an analysis of variance. Diagnostics for residuals can be
obtained. (INSIGHT will do regression interactively, except for the model selection.)
The syntax is
proc reg data=dataset;
output out=new dataset keyword(s)=name(s);
For example,
proc reg data=DATASET4;
model STRENGTH=TEMP HUMIDITY; * Regression with two predictors;
output out=DATANEW r=RESIDUAL p=PREDICTD;

Basic SAS Commands
13
Each predictor variable must have been defined in a datastep. For example, X*X cannot
be used (as it can in GLM) to indicate a quadratic term. Options for the model
statement include
i prints the inverse of the crossproducts matrix, X'X-1
cli confidence limits for predicting individual values
clm confidence limits for estimating the regression curve (mean)
noint suppresses the intercept term (default is to include it)
r requests an analysis of the residuals
ss1 prints the sequential sums of squares, according to the sequence of
predictors
ss2 prints the partial sums of squares for each predictor
vif prints the variance inflation factor for each parameter estimate
xpx prints the crossproducts matrix, X'X
Certain values may be produced by the output statement and appended to the dataset
(thus forming a new dataset). These are indicated by keyword above and include
p values predicted by the fitted model
r residuals from the fitted model
Model selection criteria can also be obtained from proc reg. Note, however, that the
above options cannot be used in conjuction with the selection option. The syntax is
proc reg data=dataset;
model response variable = predictor variables/selection=rsquare options;
Options include
adjrsq adjusted R-squared criterion
aic Akaike's infomation criterion
b provide parameter estimates
best=number list only the best (by R-square) models for each given number of
parameters
cp Mallow's Cp criterion
include=number consider only models containing at least the first number of
predictors
rmse root mean squared error criterion
sse sum of squared errors criterion
RSQUARE: Regression Model Fits.
This procedure calculates model fit for full and reduced models according to various
specified criteria. It is essentially the same as the selection procedure in REG. Use
proc rsquare data=dataset;

Basic SAS Commands
14
For example
proc rsquare data=DATASET4;
model STRENGTH=TEMP HUMIDITY/adjrsq cp rmse;
The model statement is the same as for proc reg (see REG). The options are also the
same, except for best.
SORT: Data Sorting.
The data are sorted alphabetically according to or by order of value for one or more
variables. This procedure is required before using the by statement in other procedures
to ensure the data are properly sorted. Use
proc sort data=dataset; by option variable … option variable;
If more than one variable is specified the dataset is ordered by the first variable first then
the second, etc. The default is to sort in ascending order but any variable name may be
proceeded by the option
descending to sort in descending order
TTEST: Student's Two Sample t-test.
This does the usual t-test to compare means from two independent samples. The
command is given by
proc ttest data=dataset; class classification variable;
var variables;
The classification variable is required and must consist of exactly two levels. To
compare the average weight of males and females in DATASET1,
proc ttest data=DATSET1; class SEX; var WT;
UNIVARIATE: Summary Statistics.
Summary statistics are calculated, including moments and percentiles. The information
is strictly univariate and no multiple variable analysis is provided, except that the analysis
can be broken down for categories by using a by statement. The syntax is
proc univariate data=dataset options; var variables;
Options include
normal provides a test of normality
plot provides a stem-leaf plot, a box-plot and a normal plot
def=5 the best definition for sample percentiles

Basic SAS Commands
15
The plots provided by UNIVARIATE are not particularly good because they are quite
small. But if a by statement is used then side-by-side box-plots will be provided on a full
page. Larger histograms can be obtained with CHART and normal plots can be produced
using normal quantiles obtained with RANK (see RANK).
VARCOMP: Variance Components Analysis.
This procedure analyzes models with random effects (i.e., with factors having random
values for levels).
proc varcomp data=dataset method=type1;
class variables;
model response variable = effects/option;
The effects are factors and interactions comprised from the class variables. There is
only one option.
fixed=number the first number effects are fixed effects, the remainder are random
4. Other Commands
BY: Subgroup Analysis.
Most SAS procedures can be performed on subgroups of the dataset if the subgroups are
defined by a classification variable and if the dataset is first sorted by that variable. Use
the statement
by option variable … option variable;
The class variables must appear in the same order as in the most recent sort procedure
and the option descending must be used or not used as in the sort procedure. For
example to get two scatterplots, according to SEX,
proc sort data=DATASET1; by SEX;
proc plot data=DATASET1; by SEX; plot WT*HT='*';
Or to obtain summary statistics according to TEMP and HUMIDITY and side-by-side
box-plots,
proc sort data=DATASET4; by TEMP descending HUMIDITY;
proc univariate data=DATASET4 plot; var STRENGTH;
by TEMP descending HUMIDITY;
LABEL: Variable Labels.
Variables may be labeled with a character string (in quotes) at any time. If the labeling
occurs in a datastep the labels will apply henceforth; otherwise they apply only to the
current procedure. The syntax is

Basic SAS Commands
16
label variable = 'character string' … variable = 'character string';
For example,
label NAME='Name of Patient' WT='Weight at Admission'
HT='Height at Admission';
OPTION: Page Options.
The statement to set up page options is
option options;
and possible options include
center center title, if any
date include job execution date on first line
ls=number line length is number characters
nocenter do not center title, if any
nodate do not include job execution date on first line
ps=number page length is number lines
RUN: Run Statement.
The run statement tells SAS that all commands for a particular procedure have been
provided and is required at the end of a sequence of commands submitted interactively.
The syntax is
run;
TITLE: Page Titles.
Page titles may be provided and changed at any time. They may have multiple lines. For
the first line of a title (given in quotes) use
title 'title';
and for additional lines use
title# 'title';
where # is the line number on which the title is to appear.

Basic SAS Commands
17
5. Operators, Expressions and Functions
OPERATORS: The usual mathematical operators +, -, *, / and ** (for exponentiation)
are available as well as the comparison operators =, >, <, >=, <=. Also, not, and and
or may be used in logical expressions.
EXPRESSIONS: Expressions are constructed according to the usual rules of syntax in
most programming languages such as C or Fortran. Logical expressions are given the
numerical values 0 (false) or 1 (true) and may be used as such.
FUNCTIONS: Standard functions are available as well as many statistical functions.
The standard functions include abs, max, min, sqrt, exp, log (natural logarithm),
and log10 (logarithm base 10). Trigonometric and hyperbolic functions include sin,
cos, tan, arsin, arcos, atan, sinh, cosh and tanh.
Function that compute sample statistics either take the form function(list) where list is a
list of values and variables (e.g., sum(1,2,X,Y,Z)) or take the form function(OF list)
where list is a list only of variables, separated by spaces, and may be less explicit (e.g.,
sum(OF X1-X10 Y1-Y5)). These functions include the following.
n no. of nonmissing arguments nmiss no. of missing arguments
sum sum mean average
min smallest value max largest value
std standard deviation var variance
skewness skewness kurtosis kurtosis
cv coefficient of variation
Among the statistical functions are those which give the cumulative probability function
for various distributions:
probnorm(Z) standard normal probt(T,DF Student's T
probchi(X,DF) chi-square probf(F,NDF,DDF) Fisher's F
probgam(X,ALPHA) gamma
The inverses to these functions (which compute percentiles) include:
probit(P) standard normal tinv(P,DF) Student's T
cinv(P,DF) chi-square finv(P,NDF,DDF) Fisher's F
gaminv(P,ALPHA) gamma
There are also functions to generate pseudo-random variables: normal(SEED),
uniform(SEED) and rangam(SEED,ALPHA). The argument SEED must be a
constant, either 0 or a 7-digit odd integer and is used only the first time the function is
evaluated. If SEED is 0 then a function of the current time is used for a seed.

Basic SAS Commands
18
6. Interactive Data Analysis (INSIGHT)
Most of the basic model fitting and testing (as in REG or GLM) can be done interactively
with SAS using INSIGHT, which is windows oriented and menu driven. It is also the
best and easiest way to get decent graphs. Before it can be used you first must create the
desired data (with a datastep or as output from a procedure).
To start INSIGHT select Analyze-Interactive Data Analysis from the Globals menu. This
brings up a menu of data libraries. Choose WORK and then the dataset desired. SAS
then provides a spreadsheet of the data. (It is also possible to enter data into the
spreadsheet.) By double clicking on a variable in the spreadsheet, you bring up a dialog
box for that variable. You can then change or assign labels, variable type and other
attributes.
Now choose the desired analysis from the Analyze menu. To select the variables for the
analysis, highlight the variable names and then click on Y or X as desired. In addition,
each analysis has a number of options concerning the type of plots to be produced. Both
DISTRIBUTION and FIT analyses include options for "smooth" fitting of the data. You
can also obtain residuals from the FIT analysis and perform residuals diagnostics.
Note that the output is graphical. Tables and graphs should be saved separately as
follows. To save tables, hold the Ctrl key down while highlighting each table you want
to save. Select File-Save-Tables, which copies the tables to the OUTPUT window. Select
the OUTPUT window, then select File-SaveAs to save the output as a text file (e.g.,
file.txt) which may be edited later. You cannot edit in the OUTPUT window. To save a
graph, first highlight it. Then select File-Save-GraphicsFile and check GIF and grey
scale (unless you really want color). Do not save it as a bitmap file because the file will
be 10-100 times larger with no improvement in quality. Give a filename (e.g., file.gif)
and click OK. The graphics file may be inserted as a picture with a suitable word
processor. Repeat the above for each graph you want to save. (Except when graphs are
presented adjacent to each other, it is recommended you save each graph separately.)
For more information, see the Help menu while in an INSIGHT window. Specifically,
Help-References and Help-Techniques have information about the analyses available in
INSIGHT.

SAS Commands

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to SAS Commands (20)

Recently uploaded (20)

SAS Commands