The GLM Procedure

Construction of Least Squares Means

To construct a least squares mean (LS-mean) for a particular level of a particular effect, construct a row vector bold upper L according to the following rules and use it in an ESTIMATE statement to compute the value of the LS-mean:

  1. Set all upper L Subscript i that correspond to covariates (continuous variables) to their mean value.

  2. Consider effects that are contained by the particular effect. (For a definition of containing, see Chapter 16, The Four Types of Estimable Functions.) Set the upper L Subscript i that correspond to levels associated with the particular level equal to 1. Set all other upper L Subscript i in these effects equal to 0.

  3. Consider the particular effect. Set the upper L Subscript i that correspond to the particular level equal to 1. Set the upper L Subscript i that correspond to other levels equal to 0.

  4. Consider the effects that contain the particular effect. If these effects are not nested within the particular effect, then set the upper L Subscript i that correspond to the particular level to 1 slash k, where k is the number of such columns. If these effects are nested within the particular effect, then set the upper L Subscript i that correspond to the particular level to 1 slash left-parenthesis k 1 k 2 right-parenthesis, where k 1 is the number of nested levels within this combination of nested effects and k 2 is the number of such combinations. For upper L Subscript i that correspond to other levels, use 0.

  5. Consider other effects that are not yet considered. For each effect that has no nested factors, set all upper L Subscript i that correspond to this effect to 1 slash j, where j is the number of levels in the effect. For each effect that has nested factors, set all upper L Subscript i that correspond to this effect to 1 slash left-parenthesis j 1 j 2 right-parenthesis, where j 1 is the number of nested levels within a particular combination of nested effects and j 2 is the number of such combinations.

The consequence of these rules is that the sum of the Xs within any classification effect is 1. This set of Xs forms a linear combination of the parameters that is checked for estimability before it is evaluated.

For example, consider the following model: 

proc glm;
   class A B C;
   model Y=A B A*B C Z;
   lsmeans A B A*B C;
run;

Assume A has 3 levels, B has 2 levels, and C has 2 levels, and assume that every combination of levels of A and B exists in the data. Assume also that Z is a continuous variable with an average of 12.5. Then the least squares means are computed by the following linear combinations of the parameter estimates:

A B A*B C
mu 1 2 3 1 2 11 12 21 22 31 32 1 2 Z
LSM( ) 1 1/3 1/3 1/3 1/2 1/2 1/6 1/6 1/6 1/6 1/6 1/6 1/2 1/2 12.5
LSM(A1) 1 1 0 0 1/2 1/2 1/2 1/2 0 0 0 0 1/2 1/2 12.5
LSM(A2) 1 0 1 0 1/2 1/2 0 0 1/2 1/2 0 0 1/2 1/2 12.5
LSM(A3) 1 0 0 1 1/2 1/2 0 0 0 0 1/2 1/2 1/2 1/2 12.5
LSM(B1) 1 1/3 1/3 1/3 1 0 1/3 0 1/3 0 1/3 0 1/2 1/2 12.5
LSM(B2) 1 1/3 1/3 1/3 0 1 0 1/3 0 1/3 0 1/3 1/2 1/2 12.5
LSM(AB11) 1 1 0 0 1 0 1 0 0 0 0 0 1/2 1/2 12.5
LSM(AB12) 1 1 0 0 0 1 0 1 0 0 0 0 1/2 1/2 12.5
LSM(AB21) 1 0 1 0 1 0 0 0 1 0 0 0 1/2 1/2 12.5
LSM(AB22) 1 0 1 0 0 1 0 0 0 1 0 0 1/2 1/2 12.5
LSM(AB31) 1 0 0 1 1 0 0 0 0 0 1 0 1/2 1/2 12.5
LSM(AB32) 1 0 0 1 0 1 0 0 0 0 0 1 1/2 1/2 12.5
LSM(C1) 1 1/3 1/3 1/3 1/2 1/2 1/6 1/6 1/6 1/6 1/6 1/6 1 0 12.5
LSM(C2) 1 1/3 1/3 1/3 1/2 1/2 1/6 1/6 1/6 1/6 1/6 1/6 0 1 12.5

Setting Covariate Values

By default, all covariate effects are set equal to their mean values for computation of standard LS-means. The AT option in the LSMEANS statement enables you to set the covariates to whatever values you consider interesting.

If any effect contains two or more covariates, the AT option sets the effect equal to the product of the individual means rather than the mean of the product (as with standard LS-means calculations). The AT MEANS option leaves covariates equal to their mean values (as with standard LS-means) and incorporates this adjustment to crossproducts of covariates.

For example, the following statements specify a model with a classification variable A and two continuous variables, x1 and x2: 

class A;
model y = A x1 x2 x1*x2;

The coefficients for the continuous effects with various AT specifications are shown in the following table.

Syntax x1 x2 x1*x2
lsmeans A; x 1 overbar x 2 overbar ModifyingAbove x 1 x 2 With bar
lsmeans A / at means; x 1 overbar x 2 overbar x 1 overbar dot x 2 overbar
lsmeans A / at x1=1.2; 1.2 x 2 overbar 1.2 dot x 2 overbar
lsmeans A / at (x1 x2)=(1.2 0.3); 1.2 0.3 1.2 dot 0.3

For the first two LSMEANS statements, the A LS-mean coefficient for x1 is x 1 overbar (the mean of x1) and for x2 is x 2 overbar (the mean of x2). However, the coefficient for x1*x2 is ModifyingAbove x 1 x 2 With bar for the first LSMEANS statement, but it is x 1 overbar dot x 2 overbar for the second LSMEANS statement. The third LSMEANS statement sets the coefficient for x1 equal to 1.2 and leaves the coefficient for x2 at x 2 overbar, and the final LSMEANS statement sets these values to 1.2 and 0.3, respectively.

The covariate means used in computing the LS-means are affected by the AT option in the LSMEANS statement as follows:

  • If you use a WEIGHT statement:

    • If you do not specify the AT option, unweighted covariate means are used for the covariate coefficients.

    • If you specify the AT option, weighted covariate means are used for the covariate coefficients for which no explicit AT values are specified (thus for all the covariate means if you specify AT MEANS).

  • If any observations have missing dependent variable values:

    • If you do not specify the AT option, these observations are not included in computing the covariate means.

    • If you specify the AT option, these observations are included unless they form a missing cell (a combination of CLASS variables all of whose responses are missing).

These conditions are summarized in Table 13. You can use the E option in conjunction with the AT option to verify that the modified LS-means coefficients are the ones you want.

Table 13: Treatment of Covariate Means with and without the AT Option

Without AT With AT
WEIGHT statement Weights not used Weights used
Observations with missing
dependent variable values 		Observations not included Observations included


The AT option is disabled if you specify the BYLEVEL option, in which case the coefficients for the covariates are set equal to their means within each level of the LS-mean effect in question.

Changing the Weighting Scheme

The standard LS-means have equal coefficients across classification effects; however, the OM option in the LSMEANS statement changes these coefficients to be proportional to those found in the input data set. This adjustment is reasonable when you want your inferences to apply to a population that is not necessarily balanced but has the margins observed in the original data set.

In computing the observed margins, PROC GLM uses all observations for which there are no missing independent variables, including those for which there are missing dependent variables. Also, if there is a WEIGHT variable, PROC GLM uses weighted margins to construct the LS-means coefficients. If the analysis data set is balanced or if you specify a simple one-way model, the LS-means will be unchanged by the OM option.

The BYLEVEL option modifies the observed-margins LS-means. Instead of computing the margins across the entire data set, PROC GLM computes separate margins for each level of the LS-mean effect in question. The resulting LS-means are actually equal to raw means in this case. The BYLEVEL option disables the AT option if it is specified.

Note that the MIXED procedure implements a more versatile form of the OM option, enabling you to specifying an alternative data set over which to compute observed margins. If you use the BYLEVEL option, too, then this data set is effectively the "population" over which the population marginal means are computed. For more information, see Chapter 84, The MIXED Procedure.

You might want to use the E option in conjunction with either the OM or BYLEVEL option to verify that the modified LS-means coefficients are the ones you want. It is possible that the modified LS-means are not estimable when the standard ones are, or vice versa.

Estimability of LS-Means

LS-means are defined as certain linear combinations of the parameters. As such, it is possible for them to be inestimable. In fact, it is possible for a pair of LS-means to be both inestimable but their difference estimable. When this happens, only the entries that correspond to the estimable difference are computed and displayed in the Diffs table. If ADJUST=SIMULATE is specified when there are inestimable LS-means differences, adjusted results for all differences are displayed as missing.

Last updated: December 09, 2022