Practical Data Management and Statistical Computing (BioEp691F)

Contacts

Outline

Assignments

Resources

---

Outline: Lec1 Lec2 Lec3 Lec4 Lec5 Lec6 Lec7 Lec8 Lec9 Lec10
Lectures: Lec1 Lec2 Lec3 Lec4 Lec5 Lec6 Lec7 Lec8 Lec9 Lec10

---

Lecture 6

---

1. Review: Problems in Reading Data from ASCII Files (from Assignment #4) Use column input to read the ASCII data set lec3a.dta into SAS (as well as you can). Do not change anything in the ASCII data set. [Note that it is not possible to read all data values in correctly.] (lec3ap1.sas) or (lec3ap2.sas)

• Notice jagged left border when Selecting data from the WEB- This indicates the records have different length.
• The MISSOVER option in the INFILE statement. (lec3ap4.sas )
• Edit the data-add some blank spaces to make lines of equal length in WORD or NOTEPAD (lec3ap5.sas which uses the data set lec3a1.dta)
• Make sure the way data is entered can be read! (before it is all entered)
• Other solutions:

Read data in LIST input mode: lec3ap6.sas

• 1. PROC CHART- creates bar charts and histograms. (read ONLINE Documentation in SAS PROCEDURES Guide)
• 2. PROC MEANS- calculates simple summary statistics. (read ONLINE Documentation in SAS PROCEDURES Guide)
• 3. PROC PLOT- plots one variable versus another. (read ONLINE Documentation in SAS PROCEDURES Guide)
• 4. PROC UNIVARIATE - calculates simple statistics (including the median) and forms stem and leaf plots, and box plots. (read ONLINE Documentation in SAS PROCEDURES Guide)
• 5. PROC GCHART (read ONLINE Documentation in SAS/Graph Software)

You can copy output to a Word Processor, and use these results in a report.

We illustrate how to obtain bar charts and histograms illustrating the SAS Procedures:

• PROC CHART;
• PROC FORMAT;
• PROC MEANS;

We use as an example data from the Intensive Care Study (icu.dat). A code book (icu.txt) describes the variables representing the columns, and is neede to interpret the results. We assume that you:

• i. have a copy of SAS installed on your computer
• ii. have a printed copy of the code book (icu.txt)
• iii. have previously saved a copy of the file icu.dat on the "c:\temp\" directory on your computer.
• iv. have opened the SAS program.
• v. have copied the program CHARTP1.SAS into the SAS program window.

Vertical and horizontal bar charts can be obtained for variables with discrete values, and histograms for continuous variables with PROC CHART. The OPTIONS statement controls the size of the figure that is produced. Other options in the PROC CHART procedure control chart appearance.

INFILE 'c:\temp\icu.dat' FIRSTOBS=11;
INPUT id sta age sex race ser can crn inf cpr sys hra
pre typ fra po2 ph pco bic cre loc;

• A more detailed description of statements reading the data into SAS is given elsewhere.

VALUE sexf

0="Male"
1="Female";

VALUE staf

0="Live"
1="Die";

DATA icu1;

SET icu;
FORMAT

sex sexf.
sta staf.;

• We substitute names for the values using PROC FORMAT. The set of names that are attached is assigned a Format Name given in a VALUE statement. For the variable sex, the format name assigned is sexf . The description of the names is given by equating the value to a descriptive name. Once a list of all values is given, the VALUE statement ends in a semi- colon. More than one VALUE statement can be used in the PROC FORMAT procedure.
• Once the PROC FORMAT procedure has been run, other procedures in SAS can use the formats that have been defined by including an optional FORMAT statement. We include this statement in a DATA Step. The DATA step creates a new SAS System data set named icu1 from the SAS data set icu. In the new SAS data set, the FORMAT statement tells the SAS program to use the format names (ie sexf ) in place of the values for the variable (ie. sex). A period must be included after the format name in a format statement ( sexf. ) for the program to distinguish the format name from the variable..

VBAR sex /DISCRETE;
TITLE2 "Figure 1. Frequency Distribution of Gender in Intensive Care Unit Study";

• These statements illustrate produce a simple vertical bar chart for the variable sex. The VBAR statement produces a vertical bar chart. A similar horizontal bar chart can be constructed using a HBAR statement in place of a VBAR statement. The option DISCRETE indicates that the values of the variable sex are discrete.

• The size of the chart can be controlled with an OPTIONS statement. The statement PAGESIZE= sets the number of lines on a page for the chart. The width of the chart can be controlled with a LINESIZE= command in the OPTIONS statement. Once an OPTIONS statement is given, the values for the options are in effect until a new OPTIONS statement is issued.

VBAR sex /DISCRETE GROUP=sta;
TITLE2 "Figure 2. Frequency Distribution of Gender by Survival Status";
TITLE3 " for subjects in the Intensive Care Unit Study";

• These statements illustrate a vertical bar chart for the variable sex, split by the variable sta (survival status). The variable used as a grouping variable is indicated by the GROUP= option.

VBAR sex /

DISCRETE
GROUP=sta
TYPE=PERCENT;

TITLE2 "Figure 3. Percent Frequency Distribution of Gender by Survival Status";
TITLE3 " for subjects in the Intensive Care Unit Study";

• These statements illustrate a vertical bar chart with the vertical axis representing percent. Note that the percent is calculated relative to the total number of subjects (in this example 200 subjects), and not as a percent for each of the levels of the GROUP variable sta.

VBAR age / SPACE=0;
TITLE2 "Figure 4. Frequency Distribution of Age for Subjects in the ";
TITLE3 " Intensive Care Unit Study";

• Histograms can be created for continuous variables such as age using PROC CHART. The option SPACE= sets the amount of space between the bars. Equal width intervals for age are automatically determined for the variable. The values of age in the histogram are the midpoints of the class intervals.

VAR age;
TITLE2 "Table 1. Summary Statistics for Age";
TITLE3 " Intensive Care Unit Study";

PROC CHART DATA=icu1;

VBAR age /

MIDPOINTS=10 TO 90 BY 10
SPACE=0;

TITLE2 "Figure 5. Frequency Distribution of Age for Subjects in the ";
TITLE3 " Intensive Care Unit Study";

• The interval midpoints can be set explicitly in a histogram using the MIDPOINTS= option in PROC CHART. First, we determine the range of values for the variable age using PROC MEANS. The variables whose statistics are evaluated are given following the VAR statement. The results from this procedure tell us that age ranges from 16 to 92 years for the subjects. Knowing the range, we define age classes from 5-15, 15-25, etc. These classes are specified in PROC CHART using the MIDPOINTS= statement.

VBAR age /

MIDPOINTS=10 TO 90 BY 10
SPACE=0
TYPE=PERCENT;

TITLE2 "Figure 6. Percent Relative Frequency Distribution of Age for Subjects in the ";
TITLE3 " Intensive Care Unit Study";

• The TYPE=PERCENT option in PROC CHART specifies that the vertical axis represent percent rather than frequency.

VBAR age /

MIDPOINTS=10 TO 90 BY 10
SPACE=0
TYPE=CPERCENT;

TITLE2 "Figure 7. Cumulative Percent Distribution of Age for Subjects in the";
TITLE3 " Intensive Care Unit Study";

RUN;

• The TYPE=CPERCENT options in PROC CHART specifies that the percent represented in the intervals are the cumulated percent of subjects have age less than or equal to the age in the interval.