PROC VARCOMP estimates the variance components of random variables. If you use the default METHOD=MIVQUE0 or METHOD=TYPE1, some estimates of the variance components can become negative. These negative estimates arise for a variety of reasons, such as the following:

• The variability in your data might be large enough to produce a negative estimate, even though the true value of the variance component is positive.
• Your data might contain outliers.
• A different model for interpreting your data might be more appropriate. Under some statistical models for variance component analysis, negative estimates are an indication that observations in your data are negatively correlated.

Due to the nature of the algorithms used for METHOD=ML and METHOD=REML, negative estimates are constrained to zero. The true REML estimates of the variance components can be obtained by adding the NOBOUND option and in some cases, those estimates can be negative.

If you are satisfied that the model PROC VARCOMP is using is appropriate for your data, it is common practice to treat negative variance components as if they were zero.