SAS Workshops

An introduction to the SAS system. This eight hour workshop concentrates on the SAS DATA STEP with emphasis on data input, manipulation, output and summary.

General Considerations

Environment

Modes of Operation

• Interactive
  • SAS — Windowed
• Non-Interactive (batch)
  • SAS fn (output files)
  • Requires Editor (write, execute and examine)

SAS Program Structure

Types of Blocks

• DATA Steps
• PROC Steps

Path of Execution

• Scan Statements
• Delimit Blocks (keywords)
• Execute Blocks Sequentially (any number, any order)

Writing Programs to Reflect SAS Structure

• Separation of Blocks
• Indentation within Blocks
• Use of Comments (human factor)
• Use of Capitalization

DATA STEP

Description of a SAS Data Set

What is in a Data Set

Difference Between a Data Set and a Data File

Getting Data In

Input Statement

• List (Free Field) Input
• lumn Input
• Formatted Input (Pointers)
• Mixed Input
• Importance of Missing Values

Data Control Vector (DCV)

• Status During Data Step Iterations
• Maintaining Rectangular Shape (missing values)

Where Data Sets Come From (sources)

• Internal Data (Cards)
• External Data (Infile, Import)
• Other Datasets
  • SET Statement
  • MERGE Statement
• PROC's
  • OUTPUT Statement

Manipulation of Data (Programming)

IF Statements

• Subsetting IF with Retention
• Subsetting IF with Deletion
• IF - THEN - ELSE (Efficiency)

DO - END Statements

• DO Groups (Multiple Statements)
• DO Loops (Iterative Calculations)
• Use of the ARRAY Statement

Controlling Output

• Implied Output
• Explicit Output (OUTPUT Statement)
• Data Set Options (KEEP, DROP, RENAME)

SAS Functions and Operators

• Handling of Missing Values
• Examples of SAS Operators & Functions

Getting Data Out

Permanent Datasets

• Libraries
• Datasets and Members

Exporting ASCII Files

• FILE Statement
• PUT Statement

Exporting Other Data Types 

SAS procedures to compute estimated sample sizes and statistical power is a two-three hour workshop. It covers both simple designs, as well as advanced designs including generalized linear mixed models.

Background Material

Basic sample size and power estimation

• Sample size for one and two sample means
• Sample size for one and two sample proportions

Advanced sample size and power estimation

• PROC GLMPOWER for fixed RCB model
• PROC GLMPOWER for literature example
  
  • Article for literature example
• PROC GLIMMIX for mixed effects normal RCB model
• PROC GLIMMIX for mixed effects binomial RCB model

Using the SAS system for setting up and running simple Generalized Linear Models and computing their associated statistical power.

• Handout
• Binomial data
• Count data
• Proportional data
• Binomial power
• Negative binomial power
• Mixed model techniques and warnings in SAS
• Stroup Agronomy journal article (2014). Rethinking the analysis of non-normal data in plant and soil science.

This covers the analysis of designed experiments with PROC MIXED. Topics include mixed models ANOVA, repeated measures, regression and spatial correlation. Contrasts, mean estimation and pair-wise comparisons using mixed models are also discussed.

Introduction

Analysis of Variance (ANOVA)

Requirements

• Data Structure
  
  • Response numeric
  • Class var's numeric or alpha
  • Yield Dataset
• Statements
  
  • CLASS Statement
  • MODEL Statement
    
    • Single terms
    • Crossed terms (*)

Additional Statements and Options

• RANDOM Statement
• REPEATED Statement
• LSMEANS Statement
  
  • DIFF Option
• CONTRAST Statement

Common Experimental Designs

• RCB
  
  • Example 1 — RCB
    
    • additional term
• Split Plot and Split Block
  
  • Example 2 — Split Plot
  • Example 3 — Split Block
• Repeated Measures
  
  • Example 4 — Repeated measures RCB

• Advanced Analyses 

• Multiple Regression
  
  • Example 1 — Meta—analysis of dairy data
• Spatial Correlation ANOVA
  
  • Example 2 — CRD with subsampling

SAS multivariate procedure are covered. The included procedures are PROC PRINCOMP, PROC CLUSTER, PROC DISCRIM, MANOVA (PROC GLM) AND PROC CANCORR.

Introduction

Example Data

Excel spreadsheet

PROC PRINCOMP

PROC CLUSTERPROC DISCRIM

MANOVA (PROC GLM)

PROC CANCORR

This covers the analysis of designed experiments with PROC ANOVA and PROC GLM. Topics include ANOVA, ANCOVA, regression, contrasts and mean estimation.

Introduction

Analysis of Variance (ANOVA)

Requirements

• Data Structure
  
  • Response numeric
  • Class var's numeric or alpha
• Statements
  
  • CLASS Statement
  • MODEL Statement
    
    • Single terms
    • Crossed terms (*)
• Handling of Missing Values

Additional Statements and Options

• MEANS Statement
  
  • Multiple Comparison Options
• LSMEANS Statement
  
  • PDIFF and STDERR Options
• CONTRAST and ESTIMATE Statements
• OUTPUT Statement
  
  • Residual analysis

Common Experimental Designs

• CRD and RCB
  
  • Example 1 — CRD
    
    • ANOVA table layout
    • - SS types
  • Example 2 — RCB
  • Additional term
• Split Plot and Split Block (Multiple Error Terms)
  
  • Example 3 — Split Plot
    
    • TEST Statement
  • Example 4 — Split Block
    • More than one TEST Statement

Regression

Requirements

• Data Structure
  
  • Response Numeric
  • Independents Numeric
• Statements
  
  • No CLASS Statement
  • MODEL Statement
    
    • Single terms
    • Crossed terms (*)
• Handling of Missing Values

Additional Statements and Options

• OUTPUT Statement
  
  • Regression Diagnostics
  • Predicted Values
• WEIGHT Statement
  
  • Variance Stabilization
• MODEL Statement Options
  
  • Predicted values (P) and CI's (CLI, CLM)
  • INT and NOINT

Regression Examples

• Simple Linear Regression (SLR)
  
  • Example 5
• Multiple Linear Regression
  • Example 6

Analysis of Covariance/Dummy Var Regression

Requirements

• Data Structure
  
  • Response Numeric
  • Covariate Numeric
  • Class Var's Numeric or Alpha
• Statements
  
  • CLASS Statement
  • MODEL Statement
    
    • Single terms
    • Crossed terms (*)
• Handling of Missing Values

Additional Statements and Options

• LSMEANS Statement
  
  • Adjusted Means in ANCOVA
• CONTRAST Statement
  
  • Comparison of means in ANCOVA
  • Comparison of slopes in D. V. R.
• MODEL Statement Options
  
  • SOLUTIONS for parameter est's in D. V. R.
  • INT and NOINT in D. V. R.

Analysis of Covariance (ANCOVA)

• Example 7
  
  • ANOVA Table

Dummy Var. Regression

• Example 8 — with INT
  
  • SOLUTIONS Option
  • Estimated Parameters for differential effect
• Example 9 — with NOINT
  
  • SOLUTIONS Option
    
    • Parameter estimates
  • CONTRAST and ESTIMATE Statements
    
    • Comparing estimated parameters and predicted values

Analysis of data utilizing regression techniques in SAS. Topics include linear and multiple regression, regression diagnostics and model comparison using dummy variables.

Introduction

Regression

Requirements

• Data Structure
  
  • Response Variables — numeric
  • Regressors — numeric
• Statements
  
  • MODEL Statement
  • MODEL Statement Options
    
    • Printing
    • Diagnostic
    • Model Specification

Regression Diagnostics

• Residual Analysis
  
  • Options: P, R, etc
  • Plotting
    
    • PLOT and PAINT Statements
• Residual Analysis (cont)
  
  • OUTPUT Statement
    
    • Summary Procedures: UNIVARIATE
    • External Plotting
• Influence and Collinearity

Additional Topics and Techniques

• Weighted Regression
• Model Comparison
  
  • PROC REG
    
    • DATA STEP => Dummy Variables
    • TEST Statement
• PROC GLM
  
  • Crossed terms (*)
  • CLASS and CONTRAST Statements

Analysis of discrete categorical data using SAS. Topics include one-way, two-way and multi-way tables, and contrasts using PROC FREQ and PROC CATMOD.

Introduction

Data

Categorical Variable Types

• Nominal
• Ordinal
• Interval

Categorical Responses

• Binary
• Summary Counts

Example Data

Categorical Procedures

PROC FREQ

• One-way contingency tables
• Two-way contingency tables
• Statements
  
  • TABLES
    
    • Options
  • WEIGHT
• Other Features
  
  • Output

PROC CATMOD

• Multi-dimensional tables
• Response Functions
• Model Types
• Linear Log-Linear
• Statements
  
  • MODEL
    
    • Options
  • WEIGHT
  • DIRECT
  • LOGLIN
• Examples
  
  • Multi-way structure
    
    • Linear
    • Log-Linear
  • One-way and two-way structures
• Contrast Statement

Analysis of discrete data using SAS. Topics include binomial, multinomial and count response types.

Introduction

PROC GENMOD

Experimental vs Observational data

• Data Types
• Binary
• Recorded Counts

Statements for Binary Data

• CLASS
• MODEL
  
  • Distribution: Binomial and Multinomial
  • Link: Logit
  • Interpretation of results
• LSMEANS
  
  • DIFF
• CONTRAST

Statements for Count Data

• CLASS
• MODEL
  
  • Distribution: Poisson
  • Link: Log
• LSMEANS
  
  • DIFF
• CONTRAST

Analysis of data through nonlinear regression models in SAS. Topics include model form, parameterization, estimation, diagnostics, comparison.

Introduction

Data Structure

Response Variable — numeric

Regressor (independent) Variable — numeric

Statements

Required

• PROC NLIN
  • METHODS
• PARMS
  
  • Parameter Definition
  • Starting Values
• MODEL
• DER (Derivatives)
  
  • Naming Convention

Optional

• BOUNDS
• OUTPUT
• Programming

Examples

Special Topics

Diagnostics

• Residual Analysis
  
  • Univariate summaries
  • Residual plots
• Nonlinearity
  
  • Profile t-plots
  • Profile Pair Sketches

Linearization

• Transformations

Parameterization

• Model Behavior
• Parameter Interpretation

Contrasts and Model Comparisons

• Hypothesis Testing
  
  • Single df tests
  • Multiple df tests

Using the SAS system for graphical display and output. Topics include charts, scatterplots, 3D surface plots and contour plots. Output of graphics to external file formats for presentation is also covered.

Introduction

Procedures

GPLOT

• Required Statements
  
  • PLOT
• Options
  
  • OVERLAY
  • VREF, HREF

GCHART

• Required Statements
  
  • VBAR, HBAR
  • PIE
• General Options
  
  • FREQ
• VBAR/HBAR Options
  
  • FREQ
  • PERCENT
  • CFREQ
  • CPERCENT
  • MEAN, SUM and SUMVAR
• PIE Options
  
  • FREQ
  • FILL

G3D

• Required Statements
  
  • PLOT
  • SCATTER
• General Options
  
  • ROTATE
  • TILT
• PLOT Options
  
  • CTOP / CBOTTOM
  • XYTYPE
• SCATTER Options
  
  • NONEEDLE
  • SHAPE

GCONTOUR

• Required Statements
  
  • PLOT
• Options
  
  • LEVELS
  • VREF / HREF

Other Statements

• SYMBOLx
• GOPTIONS
  • CBACK

Applied SAS/Graph

Regression Analysis

• Initial Graphics
  
  • GPLOT
• Linear Regression Output
  
  • GPLOT
    
    • Residual Plots
    • Observed and Predicted
    • Prediction Bounds
  • G3D
    
    • Surface plots in Multiple Regression
  • GCONTOUR
  • Surface contours in multiple regression
• Nonlinear Regression Output
  
  • GPLOT
    
    • Residual Plots
    • Observed and Predicted
    • Prediction Bounds
• DVR Output
  
  • GPLOT
  • Overlayed lines

Analysis of Variance

• Outputting Means
• Mean Bar Charts
  
  • GCHART
• Interaction Plots
  
  • GPLOT
• Categorical Analysis
• Distributional Plots
  
  • GCHART
    
    • Histograms
  • GCHART
    • Pie Charts

Saving Graphic Output

File Formats

• Bitmap
• Vector

Cleaning Up Output

• Importing and Export