The SURVEYREG Procedure

PROC SURVEYREG Statement

PROC SURVEYREG <options>;

The PROC SURVEYREG statement invokes the SURVEYREG procedure. It optionally names the input data sets and specifies the variance estimation method.

Table 2 summarizes the options available in the PROC SURVEYREG statement.

Table 2: PROC SURVEYREG Statement Options

Option	Description
ALPHA=	Sets the confidence level
DATA=	Specifies the SAS data set to be analyzed
MISSING	Treats missing values as a valid category
NAMELEN=	Specifies the length of effect names
NOMCAR	Treats missing values as not missing completely at random
ORDER=	Specifies the sort order
PLOTS=	Requests plots from ODS Graphics
RATE=	Specifies the sampling rate
TOTAL=	Specifies the total number of primary sampling units
TRUNCATE	Specifies class levels using no more than the first 16 characters of the formatted values
VARMETHOD=	Specifies the variance estimation method

You can specify the following options in the PROC SURVEYREG statement:

ALPHA=

sets the confidence level for confidence limits. The value of the ALPHA= option must be between 0 and 1, and the default value is 0.05. A confidence level of produces % confidence limits. The default of ALPHA=0.05 produces 95% confidence limits.

DATA=SAS-data-set

specifies the SAS data set to be analyzed by PROC SURVEYREG. If you omit the DATA= option, the procedure uses the most recently created SAS data set.

MISSING

treats missing values as a valid (nonmissing) category for all categorical variables, which include CLASS, STRATA, CLUSTER, and DOMAIN variables.

By default, if you do not specify the MISSING option, an observation is excluded from the analysis if it has a missing value. For more information, see the section Missing Values.

NAMELEN=n

specifies the length of effect names in tables and output data sets to be n characters, where n is a value between 40 and 200. The default length is 40 characters.

NOMCAR

treats that missing values in the variance computation as not missing completely at random (NOMCAR) for Taylor series variance estimation. When you specify this option, PROC SURVEYREG computes variance estimates by analyzing the nonmissing values as a domain or subpopulation, where the entire population includes both nonmissing and missing domains. For more information, see the section Missing Values.

By default, PROC SURVEYREG completely excludes an observation from analysis if that observation has a missing value, unless you specify the MISSING option. Note that the NOMCAR option has no effect on a classification variable when you specify the MISSING option, which treats missing values as a valid nonmissing level.

The NOMCAR option applies only to Taylor series variance estimation; it is ignored for replication methods.

ORDER=DATA | FORMATTED | FREQ | INTERNAL

specifies the sort order for the levels of the classification variables (which are specified in the CLASS statement).

This option also determines the sort order for the levels of DOMAIN variables.

This option applies to the levels for all classification variables, except when you use the (default) ORDER=FORMATTED option with numeric classification variables that have no explicit format. In that case, the levels of such variables are ordered by their internal value.

The ORDER= option can take the following values:

Value of ORDER=	Levels Sorted By
DATA	Order of appearance in the input data set
FORMATTED	External formatted value, except for numeric variables with no explicit format, which are sorted by their unformatted (internal) value
FREQ	Descending frequency count; levels with the most observations come first in the order
INTERNAL	Unformatted value

By default, ORDER=FORMATTED. For ORDER=FORMATTED and ORDER=INTERNAL, the sort order is machine-dependent.

For more information about sort order, see the chapter on the SORT procedure in the Base SAS Procedures Guide and the discussion of BY-group processing in the "Grouping Data" section of SAS Programmers Guide: Essentials.

PLOTS < ( global-plot-options ) > < = plot-request < (plot-option) > > PLOTS < ( global-plot-options ) > < = ( plot-request < (plot-option) > <…plot-request < (plot-option) >> )>

controls the plots that are produced through ODS Graphics.

When ODS Graphics is enabled and when the regression model depends on at most one continuous variable as a regressor, excluding the intercept, the PLOTS= option in the PROC SURVEYREG statement controls fit plots for the regression.

A plot-request identifies the plot, and a plot-option controls the appearance and content of the plot. You can specify plot-options in parentheses after a plot-request. A global-plot-option applies to all plots for which it is available unless it is altered by a specific plot-option. You can specify global-plot-options in parentheses after the PLOTS option.

When you specify only one plot-request, you can omit the parentheses around it. Here are a few examples of requesting plots:

plots=all
plots(weight=heatmap)=fit

When the regression model depends on at most one continuous variable as a regressor, excluding the intercept, PROC SURVEYREG provides a bubble plot or a heat map for model fitting. In a bubble plot, the bubble area is proportional to the weight of an observation. In a heat map, the heat color represents the sum of the weights at the corresponding location. The default plot depends on the number of observations in your data. That is, for a data set that contains 100 observations or less, a bubble plot is the default. For a data set that contains more than 100 observations, a heat map is the default.

ODS Graphics must be enabled before you can request a plot. For example:

ods graphics on;
proc surveyreg plots=fit;
   model height=weight;
run;

For more information about enabling and disabling ODS Graphics, see the section Enabling and Disabling ODS Graphics in Chapter 24, Statistical Graphics Using ODS.

When ODS Graphics is enabled, the ESTIMATE, LSMEANS, LSMESTIMATE, and SLICE statements can produce plots that are associated with their analyses. For information about these plots, see the corresponding sections of Chapter 20, Shared Concepts and Topics.

For general information about ODS Graphics, see Chapter 24, Statistical Graphics Using ODS.

Global Plot Options

A global-plot-option applies to all plots for which the option is available unless it is altered by a specific plot-option. You can specify the following global-plot-options:

MAXSIZE=Bmax

specifies the maximum size of the radius of the bubble plot. Any observation that maps to a larger size than Bmax is displayed at the size specified for Bmax. Bmax must be larger than the Bmin value specified in the MINSIZE= option. By default, MAXSIZE=21. This option is ignored when WEIGHT=HEATMAP.

MAXWVALUE=Wmax

displays all observations for which weight is greater than Wmax with a bubble whose radius is set by the MAXSIZE= option. By default, Wmax equals the largest weight in the data. This option is ignored when WEIGHT=HEATMAP.

MINSIZE=Bmin

specifies the minimum size of the radius of the bubble plot. Any observation that maps to a smaller size than Bmin is displayed at the size specified for Bmin. Bmin must be smaller than the Bmax value specified in the MAXSIZE= option. By default, MINSIZE=3.5. This option is ignored when WEIGHT=HEATMAP.

NBINS=nbin1 <nbin2>

specifies the number of bins for the heat map of the observation weights in the fit plot. The number of bins also depends on the value of the SHAPE= option:

If SHAPE=RECTANGULAR, the number of bins is calculated as follows:
- If you specify only one number, nbin1, then it is used for both the horizontal and vertical axes. For example, if you specify NBINS=10, then PROC SURVEYREG creates 100 bins ().
- If you specify two numbers, then nbin1 is used for the horizontal axis and nbin2 is used for the vertical axis. For example, if you specify NBINS=10 20, then PROC SURVEYREG creates 200 bins ().
If you specify SHAPE=HEXAGONAL, then PROC SURVEYREG calculates the number of rectangular bins by the method previously described. Then the procedure creates a heat map that has hexagonal bins of approximately the same size as those rectangular bins.

If you specify this option, then by default WEIGHT=HEATMAP.

If you do not specify this option, then the number of bins is determined by first using the algorithm that is discussed in the section ODS Graphics in Chapter 73, The KDE Procedure, and then multiplying the resulting numbers of bins by three.

ONLY

suppresses the default plots and requests only the plots that are specified as plot-requests.

WEIGHT=BUBBLE | HEATMAP

requests either a bubble plot or a heat map of the data as an overlay on the regression line and confidence limits band of the prediction in a fit plot. You can specify the following options:

BUBBLE: overlays a bubble plot in which the bubble area is proportional to the weight of an observation.
HEATMAP: overlays a heat map in which the heat color represents the sum of the weights at the corresponding location.

If you specify the NBINS= option, then by default WEIGHT=HEATMAP.

If you do not specify this option, then the default plot depends on the number of observations in your data. For a data set that contains 100 observations or less, the default is WEIGHT=BUBBLE. For a data set that contains more than 100 observations, the default is WEIGHT=HEATMAP.

Plot Requests

You can specify the following plot-requests:

ALL

requests all appropriate plots.

FIT < (plot-options) >

requests a plot that displays the model fitting for a model that depends on at most one regressor, excluding the intercept. The plot is either a bubble plot or a heat map that is overlaid with the regression line and confidence band of the prediction.

The FIT plot request has the following plot-options:

MAXSIZE=Bmax

MAXWVALUE=Wmax

MINSIZE=Bmin

NBINS=nbin1 <nbin2>

specifies the number of bins for the heat map of the observation weights in the fit plot. The number of bins also depends on the value of the SHAPE= option:

If SHAPE=RECTANGULAR, the number of bins is calculated as follows:
- If you specify only one number, nbin1, then it is used for both the horizontal and vertical axes. For example, if you specify NBINS=10, then PROC SURVEYREG creates 100 bins ().
- If you specify two numbers, then nbin1 is used for the horizontal axis and nbin2 is used for the vertical axis. For example, if you specify NBINS=10 20, then PROC SURVEYREG creates 200 bins ().
If you specify SHAPE=HEXAGONAL, then PROC SURVEYREG calculates the number of rectangular bins by the method previously described. Then the procedure creates a heat map that has hexagonal bins of approximately the same size as those rectangular bins.

If you specify this option, then by default WEIGHT=HEATMAP.

WEIGHT=BUBBLE | HEATMAP

requests either a bubble plot or a heat map of the data as an overlay on the regression line and confidence limits band of the prediction in a fit plot. You can specify the following options:

BUBBLE: overlays a bubble plot in which the bubble area is proportional to the weight of an observation.
HEATMAP: overlays a heat map in which the heat color represents the sum of the weights at the corresponding location.

If you specify the NBINS= option, then by default WEIGHT=HEATMAP.

SHAPE=RECTANGULAR | HEXAGONAL SHAPE=REC | HEX

requests either rectangular or hexagonal bins for a heat map of the data. Thus, this option implies WEIGHT=HEATMAP by default. By default, SHAPE=RECTANGULAR.

NONE

suppresses all plots.

RATE=value | SAS-data-set R=value | SAS-data-set

specifies the sampling rate, which PROC SURVEYREG uses to compute a finite population correction for Taylor series or bootstrap variance estimation. This option is ignored for the jackknife or balanced repeated replication (BRR) variance estimation method.

If your sample design has multiple stages, you should specify the first-stage sampling rate, which is the ratio of the number of primary sampling units (PSUs) in the sample to the total number of PSUs in the population.

You can specify the sampling rate in either of the following ways:

value: specifies a nonnegative number to use for a nonstratified design or for a stratified design that has the same sampling rate in each stratum.
SAS-data-set: specifies a SAS-data-set that contains the stratification variables and the sampling rates for a stratified design that has different sampling rates in the strata. You must provide the sampling rates in the data set variable named _RATE_. The sampling rates must be nonnegative numbers.

You can specify sampling rates as numbers between 0 and 1. Or you can specify sampling rates in percentage form as numbers between 1 and 100, which PROC SURVEYREG converts to proportions. The procedure treats the value 1 as 100% instead of 1%.

For more information, see the section Specification of Population Totals and Sampling Rates.

If you do not specify either the RATE= or TOTAL= option, the Taylor series or bootstrap variance estimation does not include a finite population correction. You cannot specify both the RATE= and TOTAL= options.

TOTAL=value | SAS-data-set N=value | SAS-data-set

specifies the total number of primary sampling units (PSUs) in the study population. PROC SURVEYREG uses this information to compute a finite population correction for Taylor series or bootstrap variance estimation. This option is ignored for the jackknife or BRR variance estimation method.

You can specify the total number of PSUs in either of the following ways:

value: specifies a positive number to use for a nonstratified design or for a stratified design that has the same population total in each stratum.
SAS-data-set: specifies a SAS-data-set that contains the stratification variables and the population totals for a stratified design that has different population totals in the strata. You must provide the stratum totals in the data set variable named _TOTAL_. The stratum totals must be positive numbers.

For more information, see the section Specification of Population Totals and Sampling Rates.

If you do not specify either the TOTAL= or RATE= option, the Taylor series or bootstrap variance estimation does not include a finite population correction. You cannot specify both the TOTAL= and RATE= options.

TRUNCATE

specifies that class levels should be determined using no more than the first 16 characters of the formatted values of the CLASS, STRATA, and CLUSTER variables.

VARMETHOD=method <(method-options)>

specifies the variance estimation method. PROC SURVEYREG provides the Taylor series method and the following replication (resampling) methods: balanced repeated replication (BRR), bootstrap, and jackknife.

Table 3 summarizes the available methods and method-options.

Table 3: Variance Estimation Methods

method	Variance Estimation Method	method-options
BOOTSTRAP	Bootstrap	CENTER=
		MH=value \| SAS-data-set
		OUTWEIGHTS=SAS-data-set
		REPS=number
		REPWARN <=NO \| YES>
		SEED=number
BRR	Balanced repeated replication	CENTER=
		FAY <=value>
		HADAMARD=SAS-data-set
		OUTWEIGHTS=SAS-data-set
		PRINTH
		REPS=number
		REPWARN <=NO \| YES>
JACKKNIFE \| JK	Jackknife	CENTER=
		OUTJKCOEFS=SAS-data-set
		OUTWEIGHTS=SAS-data-set
		REPWARN <=NO \| YES>
TAYLOR	Taylor series linearization	None

For VARMETHOD=BOOTSTRAP, VARMETHOD=BRR, and VARMETHOD=JACKKNIFE, you can specify method-options in parentheses after the variance estimation method. For example:

varmethod=BRR(reps=60 outweights=myReplicateWeights)

By default, VARMETHOD=JACKKNIFE if you also specify a REPWEIGHTS statement; otherwise, VARMETHOD=TAYLOR by default.

You can specify the following methods:

BOOTSTRAP <(method-options)>

requests variance estimation by the bootstrap method. For more information, see the section Bootstrap Method.

The bootstrap method requires at least two primary sampling units (PSUs) in each stratum for stratified designs unless you use a REPWEIGHTS statement to provide replicate weights.

You can specify the following method-options:

CENTER=FULLSAMPLE | REPLICATES

defines how to compute the deviations for the bootstrap method. You can specify the following values:

FULLSAMPLE: computes the deviations of the replicate estimates from the full sample estimate.
REPLICATES: computes the deviations of the replicate estimates from the average of the replicate estimates.

By default, CENTER=FULLSAMPLE. For more information, see the section Bootstrap Method.

MH=value | (values) | SAS-data-set

specifies the number of PSUs to select for the bootstrap replicate samples. You can provide bootstrap stratum sample sizes by specifying a list of values or a SAS-data-set. Alternatively, you can provide a single bootstrap sample size value to use for all strata or for a nonstratified design. You can specify the number of replicate samples in the REPS= option. For more information, see the section Bootstrap Method.

Each bootstrap sample size must be a positive integer and must be less than , which is the total number of PSUs in stratum h. By default, = for a stratified design. For a nonstratified design, the bootstrap sample size value must be less than n (the total number of PSUs in the sample). By default, m = n – 1 for a nonstratified design.

You can provide bootstrap sample sizes by specifying one of the following forms:

MH=value

specifies a single bootstrap sample size value to use for all strata or for a nonstratified design.

MH=(values)

specifies a list of stratum bootstrap sample size values. You can separate the values with blanks or commas, and you must enclose the list of values in parentheses. The number of values must not be less than the number of strata in the DATA= input data set.

Each stratum sample size value must be a positive integer and must be less than the total number of PSUs in the corresponding stratum.

MH=SAS-data-set

names a SAS-data-set that contains the stratum bootstrap sample sizes. You must provide the sample sizes in a data set variable named _NSIZE_ or SampleSize.

The SAS-data-set must contain all stratification variables that you specify in the STRATA statement. It must also contain all stratum levels that appear in the DATA= input data set. If formats are associated with the STRATA variables, the formats must be consistent in the two data sets.

Each value of the _NSIZE_ or SampleSize variable must be a positive integer and must be less than the total number of PSUs in the corresponding stratum.

OUTWEIGHTS=SAS-data-set

names a SAS-data-set in which to store the bootstrap replicate weights that PROC SURVEYREG creates. For information about replicate weights, see the section Bootstrap Method. For information about the contents of the OUTWEIGHTS= data set, see the section Replicate Weights Output Data Set.

This method-option is not available when you provide replicate weights in a REPWEIGHTS statement.

REPS=number

specifies the number of replicates for bootstrap variance estimation. The value of number must be an integer greater than 1. Increasing the number of replicates improves the estimation precision but also increases the computation time. By default, REPS=250.

REPWARN <=NO | YES>

controls the log message that PROC SURVEYREG provides when the parameters cannot be estimated from one or more replicates because they are not estimable in those replicates. The procedure displays this log message as a warning if you specify REPWARN or REPWARN=YES. By default, the procedure displays the log message as a note, as if you specified REPWARN=NO.

For more information, see the subsection Variance Estimation in the section Bootstrap Method.

SEED=number

specifies the initial seed for random number generation for bootstrap replicate sampling.

If you do not specify this option or if you specify a number that is negative or 0, PROC SURVEYREG uses the time of day from the system clock to obtain an initial seed.

To reproduce the same bootstrap replicate weights and the same analysis in a subsequent execution of PROC SURVEYREG, you can specify the same initial seed that was used in the original analysis.

PROC SURVEYREG displays the value of the initial seed in the "Variance Estimation" table.

BRR <(method-options)>

requests variance estimation by balanced repeated replication (BRR). This method requires a stratified sample design where each stratum contains two primary sampling units (PSUs). When you specify this method, you must also specify a STRATA statement unless you provide replicate weights by using the REPWEIGHTS statement. For more information, see the section Balanced Repeated Replication (BRR) Method.

You can specify the following method-options:

CENTER=FULLSAMPLE | REPLICATES

defines how to compute the deviations for the bootstrap method. You can specify the following values:

FULLSAMPLE: computes the deviations of the replicate estimates from the full sample estimate.
REPLICATES: computes the deviations of the replicate estimates from the average of the replicate estimates.

By default, CENTER=FULLSAMPLE. For more information, see the section Balanced Repeated Replication (BRR) Method.

FAY <=value>

requests Fay’s method, which is a modification of the BRR method. For more information, see the section Fay’s BRR Method.

You can specify the value of the Fay coefficient, which is used in converting the original sampling weights to replicate weights. The Fay coefficient must be a nonnegative number less than 1. By default, the Fay coefficient is 0.5.

HADAMARD=SAS-data-set H=SAS-data-set

names a SAS-data-set that contains the Hadamard matrix for BRR replicate construction. If you do not specify this method-option, PROC SURVEYREG generates an appropriate Hadamard matrix for replicate construction. For more information, see the sections Balanced Repeated Replication (BRR) Method and Hadamard Matrix.

If a Hadamard matrix of a particular dimension exists, it is not necessarily unique. Therefore, if you want to use a specific Hadamard matrix, you must provide the matrix as a SAS-data-set in this method-option.

In this SAS-data-set, each variable corresponds to a column and each observation corresponds to a row of the Hadamard matrix. You can use any variable names in this data set. All values in the data set must equal either 1 or –1. You must ensure that the matrix you provide is indeed a Hadamard matrix—that is, , where is the Hadamard matrix of dimension R and is an identity matrix. PROC SURVEYREG does not check the validity of the Hadamard matrix that you provide.

The SAS-data-set must contain at least H variables, where H denotes the number of first-stage strata in your design. If the data set contains more than H variables, PROC SURVEYREG uses only the first H variables. Similarly, this data set must contain at least H observations.

If you do not specify the REPS= method-option, the number of replicates is assumed to be the number of observations in the SAS-data-set. If you specify the number of replicates—for example, REPS=nreps—the first nreps observations in the SAS-data-set are used to construct the replicates.

You can specify the PRINTH method-option to display the Hadamard matrix that PROC SURVEYREG uses to construct replicates for BRR.

OUTWEIGHTS=SAS-data-set

names a SAS-data-set in which to store the replicate weights that PROC SURVEYREG creates for BRR variance estimation. For information about replicate weights, see the section Balanced Repeated Replication (BRR) Method. For information about the contents of the OUTWEIGHTS= data set, see the section Replicate Weights Output Data Set.

This method-option is not available when you provide replicate weights in a REPWEIGHTS statement.

PRINTH

displays the Hadamard matrix that PROC SURVEYREG uses to construct replicates for BRR variance estimation. When you provide the Hadamard matrix in the HADAMARD= method-option, PROC SURVEYREG displays only the rows and columns that are actually used to construct replicates. For more information, see the sections Balanced Repeated Replication (BRR) Method and Hadamard Matrix.

The PRINTH method-option is not available when you provide replicate weights in a REPWEIGHTS statement because the procedure does not use a Hadamard matrix in this case.

REPS=number

specifies the number of replicates for BRR variance estimation. The value of number must be an integer greater than 1.

If you do not use the HADAMARD= method-option to provide a Hadamard matrix, the number of replicates should be greater than the number of strata and should be a multiple of 4. For more information, see the section Balanced Repeated Replication (BRR) Method. If PROC SURVEYREG cannot construct a Hadamard matrix for the REPS= value that you specify, the value is increased until a Hadamard matrix of that dimension can be constructed. Therefore, the actual number of replicates that PROC SURVEYREG uses might be larger than number.

If you use the HADAMARD= method-option to provide a Hadamard matrix, the value of number must not be greater than the number of rows in the Hadamard matrix. If you provide a Hadamard matrix and do not specify the REPS= method-option, the number of replicates is the number of rows in the Hadamard matrix.

If you do not specify the REPS= or the HADAMARD= method-option and do not use a REPWEIGHTS statement, the number of replicates is the smallest multiple of 4 that is greater than the number of strata.

If you use a REPWEIGHTS statement to provide replicate weights, PROC SURVEYREG does not use the REPS= method-option; the number of replicates is the number of REPWEIGHTS variables.

REPWARN <=NO | YES>

For more information, see the section Balanced Repeated Replication (BRR) Method.

JACKKNIFE <(method-options)> JK <(method-options)>

requests variance estimation by the delete-1 jackknife method. For more information, see the section Jackknife Method. If you use a REPWEIGHTS statement to provide replicate weights, VARMETHOD=JACKKNIFE is the default variance estimation method.

The delete-1 jackknife method requires at least two primary sampling units (PSUs) in each stratum for stratified designs unless you use a REPWEIGHTS statement to provide replicate weights.

You can specify the following method-options:

CENTER=FULLSAMPLE | REPLICATES

defines how to compute the deviations for the bootstrap method. You can specify the following values:

FULLSAMPLE: computes the deviations of the replicate estimates from the full sample estimate.
REPLICATES: computes the deviations of the replicate estimates from the average of the replicate estimates.

By default, CENTER=FULLSAMPLE. For more information, see the section Jackknife Method.

OUTJKCOEFS=SAS-data-set

names a SAS-data-set in which to store the jackknife coefficients. For information about jackknife coefficients, see the section Jackknife Method. For information about the contents of the OUTJKCOEFS= data set, see the section Jackknife Coefficients Output Data Set.

OUTWEIGHTS=SAS-data-set

names a SAS-data-set in which to store the replicate weights that PROC SURVEYREG creates for jackknife variance estimation. For information about replicate weights, see the section Jackknife Method. For information about the contents of the OUTWEIGHTS= data set, see the section Replicate Weights Output Data Set.

This method-option is not available when you use a REPWEIGHTS statement to provide replicate weights.

REPWARN <=NO | YES>

For more information, see the section Jackknife Method.

TAYLOR

requests Taylor series variance estimation. This is the default method if you do not specify the VARMETHOD= option or a REPWEIGHTS statement. For more information, see the section Taylor Series (Linearization).

Last updated: December 09, 2022