Example 3.2. Computing Descriptive Statistics for Specific Groups
Goal
Compute specific descriptive statistics for specific combinations of several classification variables. Omit unneeded combinations of the classification variables from the report.
Save in a data set the minimum and maximum values for specific variables for use in a later example. Save record identifiers with the results.
Report
Nutritional Information about
 Breads Available in the Region
Values Per Bread Slice,
Calories in kcal, Fiber in Grams
The MEANS
Procedure
Variable N
Mean Minimum Maximum
------------------------------------------------------
calories 124 87
.59 56.00 138.00
dietary_fiber 124 2
.01 0.00 4.80
------------------------------------------------------
Type of
Bread Variable N
Mean Minimum Maximum
----------------------------------
---------------------------------
Specialty calories 68
89.44 71.00 138.00
dietary_fiber 68
2.02 0.00 4.80
Sandwich calories 56
85.34 56.00 111.00
dietary_fiber 56
1.99 0.50 3.90
----------------------------------
---------------------------------
Primary Flour
Ingredient Variable
N Mean Minimum Maximum
-----------------------------------
---------------------------------
Multigrain calories
29 88.83 76.00 108.00
dietary_fiber
29 2.37 0.00 3.90
Oatmeal calories
5 93.00 84.00 111.00
dietary_fiber
5 3.30 2.80 3.90
Rye calories
20 86.70 71.00 105.00
dietary_fiber
20 2.16 1.00 4.20
White calories
45 87.42 56.00 138.00
dietary_fiber
45 1.11 0.00 2.90
Whole Wheat calories
25 86.08 72.00 101.00
dietary_fiber
25 2.82 1.00 4.80
-----------------------------------
---------------------------------
Source Variable N
Mean Minimum Maximum
---------------------------------
--------------------------------
Bakery calories 48
92.88 71.00 138.00
dietary_fiber 48
2.10 0.30 4.40
Grocery calories 76
84.25 56.00 112.00
dietary_fiber 76
1.95 0.00 4.80
---------------------------------
--------------------------------
Type of
Source Bread Variable
N Mean Minimum Maximum
----------------------------------
--------------------------------
Bakery Specialty calories
31 93.16 71.00 138.00
dietary_fiber
31 1.99 0.30 4.40
Sandwich calories
17 92.35 73.00 111.00
dietary_fiber
17 2.31 0.50 3.90
Grocery Specialty calories
37 86.32 72.00 112.00
dietary_fiber
37 2.05 0.00 4.80
Sandwich calories
39 82.28 56.00 101.00
dietary_fiber
39 1.85 0.50 3.50
----------------------------------
--------------------------------
Source Brand
Variable N Mean Minimum
-----------------------------------
----------------------------------
Bakery Aunt Sal Bakes
calories 10 91.30 74.00
dietary_fiber 10 1.84 0.50
Demeter
calories 15 96.67 71.00
dietary_fiber 15 2.29 0.50
-----------------------------------
----------------------------------
Source Brand
Variable N Mean Minimum
-----------------------------------
----------------------------------
Bakery Downtown Bakers
calories 10 96.90 82.00
dietary_fiber 10 2.35 1.00
Pain du Prairie
calories 13 86.62 74.00
dietary_fiber 13 1.89 0.30
Grocery BBB Brands
calories 5 81.80 65.00
dietary_fiber 5 1.76 1.20
Choice 123
calories 6 80.67 71.00
dietary_fiber 6 1.52 0.50
Fabulous Breads
calories 15 82.80 71.00
dietary_fiber 15 1.84 0.00
Five Chimneys
calories 10 86.60 75.00
dietary_fiber 10 2.55 1.20
Gaia's Hearth
calories 11 89.00 77.00
dietary_fiber 11 1.83 0.00
Mill City Bakers
calories 9 85.33 66.00
dietary_fiber 9 1.92 0.50
Owasco Ovens
calories 12 83.33 72.00
dietary_fiber 12 2.20 0.80
RiseNShine Bread
calories 8 81.88 56.00
dietary_fiber 8 1.65 0.90
-----------------------------------
----------------------------------
Source Brand
Variable Maximum
-----------------------------------------------------
Bakery Aunt Sal Bakes
calories 105.00
dietary_fiber 3.90
Demeter
calories 111.00
dietary_fiber 4.20
Downtown Bakers
calories 138.00
dietary_fiber 4.30
Pain du Prairie
calories 108.00
dietary_fiber 4.40
Grocery BBB Brands
calories 90.00
dietary_fiber 2.60
Choice 123
calories 92.00
dietary_fiber 2.30
Fabulous Breads
calories 97.00
dietary_fiber 3.20
Five Chimneys
calories 98.00
dietary_fiber 3.80
Gaia's Hearth
calories 101.00
dietary_fiber 3.60
Mill City Bakers
calories 112.00
dietary_fiber 3.10
Owasco Ovens
calories 92.00
dietary_fiber 4.80
RiseNShine Bread
calories 100.00
dietary_fiber 3.10
-----------------------------------------------------
|
Example Features
| Data Set | BREAD |
| Featured Step | PROC MEANS |
| Featured Step Statements and Options | PROC MEANS statement: NONOBS option
CLASS statement
TYPES statement
|
| Formatting Features | PROC MEANS statement: MAXDEC= option; FW= option when sending output to the LISTING destination |
| Related Technique | PROC TABULATE |
A Closer Look | Viewing the Output Data Set Created by This Example
Creating Categories for Analysis with PROC MEANS
Comparing the BY Statement and CLASS Statement in PROC MEANS
Taking Advantage of the CLASS Statement in PROC MEANS
|
| ODS Enhanced Version of Related Technique | Example 6.5 |
| ODS Enhanced Version of Output Data Set from this Example | Example 6.15 |
| Other Examples That Use This Data Set | Examples 6.5 and 6.15 |
Example Overview
This example summarizes nutritional information for a sample of bread products that are available either from grocery stores or from bakeries. The goal is to identify products with the lowest calories and highest dietary fiber in several categories. The data are categorized by four variables:
| □ | source (grocery store vs. bakery) |
| □ | brand |
| □ | primary flour ingredient in the product |
| □ | type of bread (sandwich and a variety of specialty products) |
The program computes statistics only for specific combinations of the four classification variables, not for all possible combinations of the four variables. It also computes the overall statistics. The results are saved in an output data set.
The following list shows the combinations for which this example saves statistics. One request computes overall statistics, three requests compute statistics for categories defined by a single classification variable, and two requests compute statistics for the categories defined by crossing two classification variables.
| □ | overall |
| □ | type of bread |
| □ | primary flour ingredient in the product |
| □ | source |
| □ | source and type of bread |
| □ | source and brand |
PROC MEANS saves the results of the six requests in one data set. Each observation in the output data set contains the results for one category. With the exception of the request for overall statistics, this yields multiple observations per request; the request for overall statistics is saved in the output data set in one observation.
The output data set saves statistics on calories and dietary fiber and it saves information that identifies the category and request. Each observation also contains identifying information for the three breads with the lowest calories and for the three breads with the highest dietary fiber for the category that the observation represents. The identifying information for these rankings includes the brand, the primary flour ingredient, and the type. The rankings are shown in Example 6.15. See Figure 3.2b for a listing of the output data set.
Program
| |
|
| Define a format to group the types of products. Specify the missing value level, which will exist in the output data set. | value $type 'Sandwich'='Sandwich'
' '=' '
other='Specialty';
run;
|
| Compute specific statistics. | proc means data=bread n mean min max
|
| Specify the maximum number of decimal places to display the statistics. |
|
| When sending output to the LISTING destination, specify the width of the field in which to display each statistic in the output. |
|
| Suppress the column that displays the total number of observations for each category (“N Obs”). (Because there is no missing data in this data set, the “N Obs” column is identical to the “N” column. When analyzing your own data, verify this condition before applying this option.) |
|
| | title 'Nutritional Information about Breads
Available in the Region';
title2 'Values Per Bread Slice, Calories in
kcal, Fiber in Grams';
|
| Specify the variables whose values define the categories for the analysis. | class source brand flour type;
|
| Include the overall statistics in the report. |
|
| List the specific combinations of the classification variables that define the categories for the analyses. | type flour source source*type
source*brand;
var calories dietary_fiber;
|
| Create an output data set containing specific statistics. |
|
| Save the three extreme minimum calorie values for each category specified in the TYPES statement. | idgroup(min(calories) out[3]
|
| Save identifying information as well as the calorie values for the three selected observations per category. | (brand flour type calories)=
|
| Specify the variable name prefix for each of the identifying variables and for the minimum calorie value. | wherecal flourcal typecal mincal)
|
| Save the three extreme maximum fiber values for each category specified in the TYPES statement. | idgroup(max(dietary_fiber) out[3]
|
| Specify the variable name prefix for each of the identifying variables and for the maximum fiber value. | (brand flour type dietary_fiber)=
wherefiber flourfiber typefiber
maxfiber);
|
| Suppress the variable labels for the two analysis variables in the PROC MEANS output display. | label calories=' '
dietary_fiber=' ';
format type $type.;
run;
|
Related Technique
A similar report can be produced easily with PROC TABULATE, as shown in Figure 3.2a. The advantage of PROC TABULATE is that it provides you with more formatting options than PROC MEANS. The PROC MEANS step constructed six separate tables. With all six table specifications on one TABLE statement, PROC TABULATE combines the six tables into one large report. If you want PROC TABULATE to generate six separate tables, replace the single TABLE statement with a TABLE statement for each table.
Figure 3.2a. Output Produced by PROC TABULATE
Nutritional Information about Breads Available in the Region
Values Per Bread Slice, Calories in kcal, Fiber in Grams
---------------------------------------------------------------------------------
-----
| | Calories per Slice |Dietary Fiber(g) per
Slice |
|
|---------------------------+----------------------------|
| | N | Mean | Min | Max | N | Mean | Min |
Max |
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Overall |124| 87.59| 56.00| 138.00|124| 2.01| 0.00|
4.80|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Source | | | | | | | |
|
|----------------------------| | | | | | | |
|
|Bakery | 48| 92.88| 71.00| 138.00| 48| 2.10| 0.30|
4.40|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Grocery | 76| 84.25| 56.00| 112.00| 76| 1.95| 0.00|
4.80|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Primary Flour Ingredient | | | | | | | |
|
|----------------------------| | | | | | | |
|
|Multigrain | 29| 88.83| 76.00| 108.00| 29| 2.37| 0.00|
3.90|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Oatmeal | 5| 93.00| 84.00| 111.00| 5| 3.30| 2.80|
3.90|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Rye | 20| 86.70| 71.00| 105.00| 20| 2.16| 1.00|
4.20|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|White | 45| 87.42| 56.00| 138.00| 45| 1.11| 0.00|
2.90|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Whole Wheat | 25| 86.08| 72.00| 101.00| 25| 2.82| 1.00|
4.80|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Type of Bread | | | | | | | |
|
|----------------------------| | | | | | | |
|
|Specialty | 68| 89.44| 71.00| 138.00| 68| 2.02| 0.00|
4.80|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Sandwich | 56| 85.34| 56.00| 111.00| 56| 1.99| 0.50|
3.90|
|----------------------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Source |Type of Bread | | | | | | | |
|
|-------------+--------------| | | | | | | |
|
|Bakery |Specialty | 31| 93.16| 71.00| 138.00| 31| 1.99| 0.30|
4.40|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Sandwich | 17| 92.35| 73.00| 111.00| 17| 2.31| 0.50|
3.90|
|-------------+--------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Grocery |Specialty | 37| 86.32| 72.00| 112.00| 37| 2.05| 0.00|
4.80|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Sandwich | 39| 82.28| 56.00| 101.00| 39| 1.85| 0.50|
3.50|
|-------------+--------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Source |Brand | | | | | | | |
|
|-------------+--------------| | | | | | | |
|
|Bakery |Aunt Sal Bakes| 10| 91.30| 74.00| 105.00| 10| 1.84| 0.50|
3.90|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Demeter | 15| 96.67| 71.00| 111.00| 15| 2.29| 0.50|
4.20|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Downtown | | | | | | | |
|
| |Bakers | 10| 96.90| 82.00| 138.00| 10| 2.35| 1.00|
4.30|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Pain du | | | | | | | |
|
| |Prairie | 13| 86.62| 74.00| 108.00| 13| 1.89| 0.30|
4.40|
|-------------+--------------+---+-------+-------+-------+---+-------+-------+---
-----|
|Grocery |BBB Brands | 5| 81.80| 65.00| 90.00| 5| 1.76| 1.20|
2.60|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Choice 123 | 6| 80.67| 71.00| 92.00| 6| 1.52| 0.50|
2.30|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Fabulous | | | | | | | |
|
| |Breads | 15| 82.80| 71.00| 97.00| 15| 1.84| 0.00|
3.20|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Five Chimneys | 10| 86.60| 75.00| 98.00| 10| 2.55| 1.20|
3.80|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Gaia's Hearth | 11| 89.00| 77.00| 101.00| 11| 1.83| 0.00|
3.60|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Mill City | | | | | | | |
|
| |Bakers | 9| 85.33| 66.00| 112.00| 9| 1.92| 0.50|
3.10|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |Owasco Ovens | 12| 83.33| 72.00| 92.00| 12| 2.20| 0.80|
4.80|
|
|--------------+---+-------+-------+-------+---+-------+-------+--------|
| |RiseNShine | | | | | | | |
|
| |Bread | 8| 81.88| 56.00| 100.00| 8| 1.65| 0.90|
3.10|
---------------------------------------------------------------------------------
-----
|
The advantages of PROC MEANS are in the features associated with the CLASS statement, such as the TYPES and WAYS statements and their options, and in the ease with which you can create output data sets that can then be processed by other procedures. While you can create data sets with PROC TABULATE, you cannot save information about minimum and maximum values with the output data set as you can with PROC MEANS.
The following PROC TABULATE step produces the output shown in Figure 3.2a.
| | proc tabulate data=bread;
title 'Nutritional Information about Breads
Available in the Region';
title2 'Values Per Bread Slice, Calories in
kcal, Fiber in Grams';
|
| Specify the same CLASS variables as in the PROC MEANS step. | class source brand flour type;
var calories dietary_fiber;
|
| Compute an overall statistics table. |
|
| Compute the one-way statistics tables. |
|
| Compute the two-way statistics tables. | source*type source*brand,
|
| Place the analysis variables in the column dimension and nest the statistics under each variable. | (calories dietary_fiber)*
(n*f=3. (mean min max)*f=7.2) / rts=30;
format type $type.;
run;
|
| A Closer Look |
Viewing the Output Data Set Created by This Example
The statistics computed by PROC MEANS in this example are saved in an output data set. The TYPES statement specified six requests, which results in the statistics for the six requests saved in the output data set BREADSTATS.
SAS identifies each request with a unique number and saves this value in the variable _TYPE_ that it defines. (Note that the values of _TYPE_ are numeric by default; specifying the PROC MEANS statement option CHARTYPE defines these values as character values.)
Submitting the following PROC PRINT step lists the contents of the output data set BREADSTATS. The values of _TYPE_ are formatted in the report as shown in Figure 3.2b. The variables identifying items with minimum calories and highest fiber are saved in the data set.
proc format;
value _type_ 0='0: Overall'
1='1: Type'
2='2: Flour'
8='8: Source'
9='9: Source*Type'
12='12: Source*Brand';
run;
proc print data=breadstats;
title 'Nutritional Information about Breads Available in the
Region';
title2 'Values Per Bread Slice, Calories in kcal, Fiber in
Grams';
title3 'Breads with Fewest Calories and Most Dietary Fiber';
by _type_;
format _type_ _type_.;
run;Figure 3.2b shows the PROC PRINT output. The _FREQ_ variable, which is another variable that PROC MEANS defines, contains the number of observations that the given category represents.
Figure 3.2b. PROC PRINT Listing of the BREADSTATS Data Set Created by PROC MEANS
Nutritional Information about Breads Available in the Region
Values Per Bread Slice, Calories in kcal, Fiber in Grams
Breads with Fewest Calories and Most Dietary Fiber
-------------------------------------- _TYPE_=0: Overall
--------------------------------------
Obs source brand flour type _FREQ_ wherecal_1 wherecal_2 wherecal_3
1 124 RiseNShine Bread BBB Brands Mill City
Bakers
Obs flourcal_1 flourcal_2 flourcal_3 typecal_1 typecal_2 typecal_3 mincal_1 mincal_2 mincal_3
1 White White White Sandwich Sandwich Sandwich 56 65 66
Obs wherefiber_1 wherefiber_2 wherefiber_3 flourfiber_1 flourfiber_2 flourfiber_3
1 Owasco Ovens Pain du Prairie Downtown Bakers Whole Wheat Whole Wheat Whole Wheat
Obs typefiber_1 typefiber_2 typefiber_3 maxfiber_1 maxfiber_2 maxfiber_3
1 Specialty Specialty Specialty 4.8 4.4 4.3
--------------------------------------- _TYPE_=1: Type
----------------------------------------
Obs source brand flour type _FREQ_ wherecal_1 wherecal_2 wherecal_3
2 Specialty 68 Demeter Owasco Ovens Fabulous Breads
3 Sandwich 56 RiseNShine Bread BBB Brands Mill City Bakers
Obs flourcal_1 flourcal_2 flourcal_3 typecal_1 typecal_2 typecal_3 mincal_1 mincal_2
2 White Whole Wheat Whole Wheat Specialty Specialty Specialty 71 72
3 White White White Sandwich Sandwich Sandwich 56 65
Obs mincal_3 wherefiber_1 wherefiber_2 wherefiber_3 flourfiber_1
flourfiber_2
2 74 Owasco Ovens Pain du Prairie Downtown Bakers Whole Wheat Whole
Wheat
3 66 Aunt Sal Bakes Demeter Gaia's Hearth Multigrain Oatmeal
Obs flourfiber_3 typefiber_1 typefiber_2 typefiber_3 maxfiber_1 maxfiber_2
maxfiber_3
2 Whole Wheat Specialty Specialty Specialty 4.8 4.4 4.3
3 Oatmeal Sandwich Sandwich Sandwich 3.9 3.9 3.5
--------------------------------------- _TYPE_=2: Flour
---------------------------------------
Obs source brand flour type _FREQ_ wherecal_1 wherecal_2 wherecal_3
4 Multigrain 29 RiseNShine Bread Choice 123 Mill City Bakers
5 Oatmeal 5 Five Chimneys Fabulous Breads Owasco Ovens
6 Rye 20 Mill City Bakers Five Chimneys Fabulous Breads
7 White 45 RiseNShine Bread BBB Brands Mill City Bakers
8 Whole Wheat 25 Owasco Ovens Fabulous Breads Owasco Ovens
Obs flourcal_1 flourcal_2 flourcal_3 typecal_1 typecal_2 typecal_3 mincal_1
mincal_2
4 Multigrain Multigrain Multigrain Sandwich Sandwich Sandwich 76 76
5 Oatmeal Oatmeal Oatmeal Sandwich Sandwich Sandwich 84 85
6 Rye Rye Rye Sandwich Sandwich Sandwich 71 77
7 White White White Sandwich Sandwich Sandwich 56 65
8 Whole Wheat Whole Wheat Whole Wheat Specialty Specialty Sandwich 72 74
Obs mincal_3 wherefiber_1 wherefiber_2 wherefiber_3 flourfiber_1 flourfiber_2
4 77 Aunt Sal Bakes Demeter Demeter Multigrain Multigrain
5 90 Demeter Gaia's Hearth Five Chimneys Oatmeal Oatmeal
6 78 Demeter Gaia's Hearth Five Chimneys Rye Rye
7 66 Downtown Bakers Five Chimneys Aunt Sal Bakes White White
8 75 Owasco Ovens Pain du Prairie Downtown Bakers Whole Wheat Whole Wheat
Obs flourfiber_3 typefiber_1 typefiber_2 typefiber_3 maxfiber_1 maxfiber_2
maxfiber_3
4 Multigrain Sandwich Sandwich Specialty 3.9 3.5 3.3
5 Oatmeal Sandwich Sandwich Sandwich 3.9 3.5 3.3
6 Rye Specialty Specialty Specialty 4.2 3.6 3.5
7 White Specialty Sandwich Specialty 2.9 2.3 2.3
8 Whole Wheat Specialty Specialty Specialty 4.8 4.4 4.3
-------------------------------------- _TYPE_=8: Source
---------------------------------------
Obs source brand flour type _FREQ_ wherecal_1 wherecal_2 wherecal_3
9 Bakery 48 Demeter Demeter Pain du Prairie
10 Grocery 76 RiseNShine Bread BBB Brands Mill City Bakers
Obs flourcal_1 flourcal_2 flourcal_3 typecal_1 typecal_2 typecal_3 mincal_1 mincal_2
9 White White White Specialty Sandwich Specialty 71 73
10 White White White Sandwich Sandwich Sandwich 56 65
Obs mincal_3 wherefiber_1 wherefiber_2 wherefiber_3 flourfiber_1
flourfiber_2
9 74 Pain du Prairie Downtown Bakers Demeter Whole Wheat Whole Wheat
10 66 Owasco Ovens Owasco Ovens Five Chimneys Whole Wheat Whole Wheat
Obs flourfiber_3 typefiber_1 typefiber_2 typefiber_3 maxfiber_1 maxfiber_2
maxfiber_3
9 Rye Specialty Specialty Specialty 4.4 4.3 4.2
10 Whole Wheat Specialty Specialty Specialty 4.8 4.0 3.8
------------------------------------ _TYPE_=9: Source*Type
------------------------------------
Obs source brand flour type _FREQ_ wherecal_1 wherecal_2 wherecal_3
11 Bakery Specialty 31 Demeter Pain du Prairie Aunt Sal Bakes
12 Bakery Sandwich 17 Demeter Pain du Prairie Aunt Sal Bakes
13 Grocery Specialty 37 Owasco Ovens Fabulous Breads RiseNShine Bread
14 Grocery Sandwich 39 RiseNShine Bread BBB Brands Mill City Bakers
Obs flourcal_1 flourcal_2 flourcal_3 typecal_1 typecal_2 typecal_3 mincal_1 mincal_2
11 White White White Specialty Specialty Specialty 71 74
12 White Whole Wheat Rye Sandwich Sandwich Sandwich 73 80
13 Whole Wheat Whole Wheat White Specialty Specialty Specialty 72 74
14 White White White Sandwich Sandwich Sandwich 56 65
Obs mincal_3 wherefiber_1 wherefiber_2 wherefiber_3 flourfiber_1 flourfiber_2
11 74 Pain du Prairie Downtown Bakers Demeter Whole Wheat Whole Wheat
12 81 Aunt Sal Bakes Demeter Demeter Multigrain Oatmeal
13 74 Owasco Ovens Owasco Ovens Five Chimneys Whole Wheat Whole Wheat
14 66 Gaia's Hearth Five Chimneys Fabulous Breads Oatmeal Oatmeal
Obs flourfiber_3 typefiber_1 typefiber_2 typefiber_3 maxfiber_1 maxfiber_2
maxfiber_3
11 Rye Specialty Specialty Specialty 4.4 4.3 4.2
12 Multigrain Sandwich Sandwich Sandwich 3.9 3.9 3.5
13 Whole Wheat Specialty Specialty Specialty 4.8 4.0 3.8
14 Whole Wheat Sandwich Sandwich Sandwich 3.5 3.3 3.1
----------------------------------- _TYPE_=12: Source*Brand
-----------------------------------
Obs source brand flour type _FREQ_ wherecal_1 wherecal_2
15 Bakery Aunt Sal Bakes 10 Aunt Sal Bakes Aunt Sal Bakes
16 Bakery Demeter 15 Demeter Demeter
17 Bakery Downtown Bakers 10 Downtown Bakers Downtown Bakers
18 Bakery Pain du Prairie 13 Pain du Prairie Pain du Prairie
19 Grocery BBB Brands 5 BBB Brands BBB Brands
20 Grocery Choice 123 6 Choice 123 Choice 123
21 Grocery Fabulous Breads 15 Fabulous Breads Fabulous Breads
22 Grocery Five Chimneys 10 Five Chimneys Five Chimneys
23 Grocery Gaia's Hearth 11 Gaia's Hearth Gaia's Hearth
24 Grocery Mill City Bakers 9 Mill City Bakers Mill City Bakers
25 Grocery Owasco Ovens 12 Owasco Ovens Owasco Ovens
26 Grocery RiseNShine Bread 8 RiseNShine Bread RiseNShine Bread
Obs wherecal_3 flourcal_1 flourcal_2 flourcal_3 typecal_1 typecal_2 typecal_3
mincal_1
15 Aunt Sal Bakes White Whole Wheat Rye Specialty Specialty Sandwich 74
16 Demeter White White Rye Specialty Sandwich Sandwich 71
17 Downtown Bakers Rye White Whole Wheat Sandwich Specialty Specialty 82
18 Pain du Prairie White Whole Wheat Whole Wheat Specialty Specialty Sandwich 74
19 BBB Brands White Rye Rye Sandwich Sandwich Specialty 65
20 Choice 123 White White Multigrain Sandwich Sandwich Sandwich 71
21 Fabulous Breads White Whole Wheat White Sandwich Specialty Specialty 71
22 Five Chimneys White Rye White Specialty Sandwich Specialty 75
23 Gaia's Hearth White Whole Wheat Multigrain Specialty Specialty Specialty 77
24 Mill City Bakers White Rye Multigrain Sandwich Sandwich Sandwich 66
25 Owasco Ovens Whole Wheat Whole Wheat Whole Wheat Specialty Sandwich Specialty 72
26 RiseNShine Bread White White Multigrain Sandwich Specialty Sandwich 56
Obs mincal_2 mincal_3 wherefiber_1 wherefiber_2 wherefiber_3 flourfiber_1
15 81 81 Aunt Sal Bakes Aunt Sal Bakes Aunt Sal Bakes Multigrain
16 73 94 Demeter Demeter Demeter Rye
17 82 85 Downtown Bakers Downtown Bakers Downtown Bakers Whole Wheat
18 76 80 Pain du Prairie Pain du Prairie Pain du Prairie Whole Wheat
19 82 84 BBB Brands BBB Brands BBB Brands Whole Wheat
20 71 76 Choice 123 Choice 123 Choice 123 Multigrain
21 74 77 Fabulous Breads Fabulous Breads Fabulous Breads Multigrain
22 77 82 Five Chimneys Five Chimneys Five Chimneys Whole Wheat
23 80 81 Gaia's Hearth Gaia's Hearth Gaia's Hearth Rye
24 71 77 Mill City Bakers Mill City Bakers Mill City Bakers Multigrain
25 75 79 Owasco Ovens Owasco Ovens Owasco Ovens Whole Wheat
26 74 76 RiseNShine Bread RiseNShine Bread RiseNShine Bread Multigrain
Obs flourfiber_2 flourfiber_3 typefiber_1 typefiber_2 typefiber_3 maxfiber_1 maxfiber_2
maxfiber_3
15 Rye Whole Wheat Sandwich Specialty Sandwich 3.9 3.5 2.4
16 Whole Wheat Oatmeal Specialty Specialty Sandwich 4.2 3.9 3.9
17 Whole Wheat Rye Specialty Sandwich Specialty 4.3 3.0 3.0
18 Whole Wheat Whole Wheat Specialty Sandwich Specialty 4.4 2.9 2.8
19 Rye Rye Sandwich Specialty Sandwich 2.6 2.0 1.5
20 Multigrain Whole Wheat Specialty Sandwich Sandwich 2.3 2.0 1.8
21 Whole Wheat Rye Specialty Sandwich Specialty 3.2 3.1 3.1
22 Rye Oatmeal Specialty Specialty Sandwich 3.8 3.5 3.3
23 Oatmeal Whole Wheat Specialty Sandwich Sandwich 3.6 3.5 2.3
24 Multigrain Whole Wheat Specialty Sandwich Sandwich 3.1 2.9 2.6
25 Whole Wheat Rye Specialty Specialty Specialty 4.8 4.0 2.9
26 Whole Wheat White Specialty Sandwich Specialty 3.1 2.0 1.7
|
Creating Categories for Analysis with PROC MEANS
PROC MEANS provides several ways to define categories in generating statistics tables. The remaining topics in this section describe some of the ways you can create categories in PROC MEANS.
When you save PROC MEANS results in a data set, you can then go on to produce customized reports using other SAS procedures, DATA steps, and ODS. Understanding your choices for defining categories for analysis will help you more efficiently use PROC MEANS to create data sets that can in turn be used to generate your final reports.
Example 6.15 uses DATA steps and ODS features to produce a report summarizing some of the information saved in the output data set created by PROC MEANS in this example.
Comparing the BY Statement and CLASS Statement in PROC MEANS
The two main ways to tell PROC MEANS you want your analyses done by categories are the BY statement and the CLASS statement.
When you use a BY statement, you request a separate analysis of each BY group. Your analysis data set must be sorted or indexed by the variables on the BY statement, or you must add the NOTSORTED option to the BY statement in the PROC MEANS step.
When you use the CLASS statement, you specify the variables whose values define the categories for the analysis. If the PROC MEANS step includes no other statements or options associated with the CLASS statement, the categories are defined by all the possible combinations of the CLASS variables, with all CLASS variables represented in each category. For example, if you have four CLASS variables, the categories for analysis are all the four-way combinations of the values of the four variables.
You do not have to sort or index your analysis data set by the CLASS variables before executing a PROC MEANS step that contains a CLASS statement, but no BY statement.
If you have several CLASS variables or your CLASS variables have many values, your report can be quite lengthy and may require additional computing resources because of the complexity. PROC MEANS must keep a copy of each unique value of each CLASS variable in memory. So in situations of limited computing resources, you may want to change some of your CLASS variables to BY variables. Also, you can adjust the PROC MEANS option SUMSIZE= to provide potentially more memory for your PROC MEANS step.
You can use a BY statement and a CLASS statement in the same PROC MEANS step. When you do, SAS analyzes the data by each BY group and applies the CLASS statement to each BY group.
Taking Advantage of the CLASS Statement in PROC MEANS
Several statements and options can be used in conjuction with the CLASS statement. These features allow you to produce tables of statistics for specific combinations of the CLASS variables as well as for specific values of the CLASS variables. You can also create output data sets that contain only these specific tables.
Understanding the choices in how to define categories will help you more efficiently focus only on the tables and categories that you need.
The BREAD data set contains two values for SOURCE, twelve values for BRAND, two values for TYPE when the TYPE format has been applied, and five values for FLOUR. To obtain the statistics shown in the preceding PROC MEANS report, without using the TYPES statement, add the PRINTALLTYPES to the PROC MEANS statement, and delete the TYPES statement. A modified version of the featured PROC MEANS step that does this follows:
proc means data=bread n mean min max maxdec=2 fw=7
nonobs printalltypes;
title 'Nutritional Information about Breads Available
in the Region';
title2 'Values Per Bread Slice, Calories in kcal,
Fiber in Grams';
class source brand flour type;
var calories dietary_fiber;
output out=breadstats
idgroup(min(calories) out[3]
(brand flour type calories)=
wherecal flourcal typecal mincal)
idgroup(max(dietary_fiber) out[3]
(brand flour type dietary_fiber)=
wherefiber flourfiber typefiber maxfiber);
label calories=' '
dietary_fiber=' ';
format type $type.;
run;The PROC MEANS step above produces tables for all one-way, two-way, three-way, and four-way combinations of the values of the four class variables. It also produces an overall table. This yields sixteen tables when the original request as specified on the TYPES statement specified six: five requests with classification variables plus one overall table. The output data set created by the PROC MEANS step that computes all sixteen tables contains 405 observations, whereas the output data set from the PROC MEANS step that computes the six specific tables contains 26 observations.
The four-way table alone in the PROC MEANS step that includes the PRINTALLTYPES option could generate statistics for a maximum of 240 categories: 2 sources*12 brands*2 formatted values of type*5 flours=240. (The BREAD data set, however, does not contain data for all combinations. The total number of categories in the four-way table for this data set is 89.)
Table 3.2a illustrates several ways to define categories when you use the CLASS statement in combination with other PROC MEANS features.
Table 3.2a. Combining the CLASS Statement with PROC MEANS to Define Categories| Purpose | Specification | Description and Example |
|---|
| Produce a table of statistics for the most complete combination of the CLASS variables | Additional statement: none
PROC MEANS statement options: none
CLASS statement options: none | proc means data=bread;
class source brand flour type;
var calories dietary_fiber;
run;
Result: Table with statistics for each combination of the values of the four variables taken four variables at a time.
Restrictions: The one-, two-, and three-way combinations are not evaluated.
If many CLASS variables exist and/or the CLASS variables have many values, the report can be lengthy and can require additional computing resources.
|
| Produce tables of statistics for all possible combinations of the CLASS variables at all levels, plus an overall table | Additional statement: none
PROC MEANS statement options: PRINTALLTYPES
CLASS statement options: none | proc means data=bread printalltypes;
class source brand flour type;
var calories dietary_fiber;
run;
Result: Tables with statistics for each combination of the values of the four variables taken one variable at a time, two variables a time, three variables at a time, and four variables at a time. Also includes a table of overall statistics.
Restrictions: If many CLASS variables exist and/or the CLASS variables have many values, the report can be lengthy and can require additional computing resources.
|
| Produce tables for selected combinations of the CLASS variables | Additional statement: TYPES
PROC MEANS statement options: none
CLASS statement options: none | proc means data=bread;
class source brand flour type;
types () type flour source source*type
source*brand;
var calories dietary_fiber;
run;
Result: Six tables of statistics: overall; three 1-way, and two 2-way. The one-way tables: categories defined by TYPE; by FLOUR; and by SOURCE. The two-way tables: categories defined by the combinations of SOURCE and TYPE; and by SOURCE and BRAND.
Note: The text () specifies the computation of an overall summary.
|
| Produce tables for selected levels of combinations of the CLASS variables | Additional statement: WAYS
PROC MEANS statement options: none
CLASS statement options: none | proc means data=bread;
class source brand flour type;
ways 2 4;
var calories dietary_fiber;
run;
Result: Seven tables of statistics: six 2-way tables and one 4-way table. The 2-way tables: categories defined by the combination of SOURCE and BRAND; SOURCE and FLOUR; SOURCE and TYPE; BRAND and FLOUR; BRAND and TYPE; and FLOUR and TYPE. The 4-way table: categories defined by the combination of the values of SOURCE, BRAND, FLOUR, and TYPE.
Restrictions: If many CLASS variables exist and/or the CLASS variables have many values, the report can be lengthy and can require additional computing resources.
|
| Produce a table of statistics for the most complete combination of the CLASS variables. Included in the output are the categories defined in the CLASSDATA= data set, even if the categories are not present in the analysis data set. These categories have an N of 0. | Additional statement: None
PROC MEANS statement options: CLASSDATA=
CLASS statement options: none | proc means data=bread;
class source brand flour type;
ways 2 4;
var calories dietary_fiber;
run;
PROC PRINT of data set COMPLEX:
Obs source flour
1 Grocery Whole Wheat
2 Bakery Whole Wheat
3 Other Whole Wheat
4 Grocery Multigrain
5 Bakery Multigrain
6 Other Multigrain
Result: Twelve categories in the output. One table that contains all combinations of the values of SOURCE and FLOUR that are in the BREAD data set plus two additional rows. These two rows are SOURCE=’Other’ and FLOUR=’Whole Wheat’, and SOURCE=’Other’ and FLOUR=’Multigrain’. These two rows have an N of 0, because of no representation in the BREAD data set.
Restrictions: Must include all CLASS variables in the CLASSDATA= data set.
CLASS variables in the CLASSDATA= data set must be defined exactly as they are in the analysis data set.
|
| Produce a table of statistics for all possible combinations of CLASS variables, including those in the analysis data set as well as those represented in the CLASSDATA= data set, even if the categories are not present in the analysis data set. Categories without representation in the analysis data set have an N of 0. | Additional statement: none
PROC MEANS statement options: CLASSDATA=, COMPLETETYPES
CLASS statement options: none | proc means data=bread classdata=complex
completetypes;
class source flour;
var calories dietary_fiber;
run;
This example uses the same COMPLEX data set as listed in the previous row in this table.
Result: Fifteen categories in the output. One table that contains all the combinations of the values of SOURCE and FLOUR that are present in the BREAD data set and in the COMPLEX data set (3 values of SOURCE and 5 values of FLOUR=15 categories). The five rows where SOURCE=’Other’ have an N of 0.
Restrictions: Must include all CLASS variables in the CLASSDATA= data set.
CLASS variables in the CLASSDATA= data set must be defined exactly as they are in the analysis data set.
If specifications in the CLASSDATA= data set are not in the analysis data set, there are potentially many rows in the table with no statistics other than N=0. This example has five rows with N=0.
|
| Produce a table of statistics for the most complete combination of the CLASS variables for which categories are defined in the CLASSDATA= data set, even if the categories are not present in the analysis data set. These categories have an N of 0. Do not include categories that are only in the analysis data set and not in the CLASSDATA= data set. | Additional statement: none
PROC MEANS statement options: CLASSDATA=, EXCLUSIVE
CLASS statement options: none | proc means data=bread classdata=complex
exclusive;
class source flour;
var calories dietary_fiber;
run;
This example uses the same COMPLEX data set as used in the previous two rows of this table.
Result: Six categories in one table in the output. The table contains only the combinations of the values of SOURCE and TYPE that are present in the CLASSDATA= data set. Combinations that are present only in the analysis data set are excluded from the results. Two categories are generated for each of the three values of SOURCE that are present in COMPLEX, one for TYPE=’Whole Wheat’ and one for TYPE=’Multigrain’. The two rows for SOURCE=’Other’ have an N of 0.
Restrictions: Must include all CLASS variables in the CLASSDATA= data set.
CLASS variables in the CLASSDATA= data set must be defined exactly as they are in the analysis data set.
|
| Produce tables for all combinations of the CLASS variables as specified by the formats associated with them, even if the combinations are not in the analysis data set | Additional statement: FORMAT, if formats not already assigned to CLASS variables
PROC MEANS statement options: COMPLETETYPES
CLASS statement options: PRELOADFMT | proc format;
value $texture
'Multigrain','Oatmeal','Whole
Wheat'='Whole Grain'
'White','Bleached White'='Refined'
'Garbanzo','Soy'='Beans';
run;
proc means data=bread completetypes;
class flour / preloadfmt;
var calories dietary_fiber;
format flour $texture.;
run;
Result: Four categories in one table in the output. The table contains the categories in the variable FLOUR as formatted by $TEXTURE. Values not represented in $TEXTURE are included in the analysis as their unformatted values. Values that are in $TEXTURE, but not in the BREAD data set, are included in the analysis. The report will show a row for “Beans” with N=0, because there are no products in the BREAD data set whose primary flour ingredient is garbanzo bean flour or soy flour. The report will show a row for “Rye,” the one value in the BREAD data set that is not included in $TEXTURE.
Restrictions: Need to define formats for the CLASS variables before the PROC MEANS step.
|
| Produce tables for all combinations of the CLASS variables as specified by the formats associated with them. Omit combinations that are in the associated format, but not in the data set. Omit combinations that are in the data set and not in the associated format. | Additional statement: FORMAT, if formats not already assigned to CLASS variables
PROC MEANS statement options: none
CLASS statement options: PRELOADFMT, EXCLUSIVE | proc format;
value $texture
'Multigrain','Oatmeal','Whole
Wheat'='Whole Grain'
'White','Bleached White'='Refined'
'Garbanzo','Soy'='Beans';
run;
proc means data=bread;
class flour / preloadfmt exclusive;
var calories dietary_fiber;
format type $texture.;
run;
Result: One table that contains the categories in the variable FLOUR as formatted by $TEXTURE and restricted by the EXCLUSIVE option: one category for “Whole Grain” and one for “Refined.” The one value of FLOUR not represented in $TEXTURE ( “Rye”) is not included in the report. The one formatted value in $TEXTURE (“Beans”) that does not have any corresponding values in FLOUR is not included in the report.
Restrictions: Need to define formats for the CLASS variables before the PROC MEANS step.
Note: This discussion places the EXCLUSIVE option on the CLASS statement. If you place the EXCLUSIVE option on the PROC MEANS statement instead, your report will contain the same results plus a row for “RYE.”
|
Table 3.2a does not describe the process of defining categories in which some of the values are represented multiple times. You can do this by creating multilabel formats with PROC FORMAT. See Example 3.7 for an example of applying multilabel formats.
Where to Go from Here
PROC FORMAT reference, usage information, and additional examples. See “The FORMAT Procedure” in the “Procedures” section of Base SAS 9.1 Procedures Guide.
PROC MEANS reference, usage information, and additional examples. See “The MEANS Procedure” in the “Procedures” section of Base SAS 9.1 Procedures Guide.
PROC TABULATE reference, usage information, and additional examples. See, “The TABULATE Procedure” in the “Procedures” section of Base SAS 9.1 Procedures Guide.
