Example 77.2 Robust ANOVA

The classical analysis of variance (ANOVA) technique based on least squares assumes that the underlying experimental errors are normally distributed. However, data often contain outliers due to recording or other errors. In other cases, extreme responses occur when control variables in the experiments are set to extremes. It is important to distinguish these extreme points and determine whether they are outliers or important extreme cases. You can use the ROBUSTREG procedure for robust analysis of variance based on M estimation. Typically, there are no high leverage points in a well-designed experiment, so M estimation is appropriate.

The following example shows how to use the ROBUSTREG procedure for robust ANOVA.

An experiment was carried out to study the effects of two successive treatments (T1, T2) on the recovery time of mice with certain diseases. Sixteen mice were randomly assigned into four groups for the four different combinations of the treatments. The recovery times (time) were recorded (in hours) as shown in the following data set recover.

data recover;
   input  T1 $ T2 $ time @@;
   datalines;
0 0 20.2  0 0 23.9  0 0 21.9  0 0 42.4
1 0 27.2  1 0 34.0  1 0 27.4  1 0 28.5
0 1 25.9  0 1 34.5  0 1 25.1  0 1 34.2
1 1 35.0  1 1 33.9  1 1 38.3  1 1 39.9
;

The following statements invoke the GLM procedure ( Chapter 41, The GLM Procedure ) for a standard ANOVA:

proc glm data=recover;
    class T1 T2;
    model time = T1 T2 T1*T2;
run;

Output 77.2.1 Overall ANOVA
The GLM Procedure
 
Dependent Variable: time

Source DF Sum of Squares Mean Square F Value Pr > F
Model 3 209.9118750 69.9706250 1.86 0.1905
Error 12 451.9225000 37.6602083    
Corrected Total 15 661.8343750      

R-Square Coeff Var Root MSE time Mean
0.317167 19.94488 6.136791 30.76875

Output 77.2.2 Model ANOVA
Source DF Type I SS Mean Square F Value Pr > F
T1 1 81.4506250 81.4506250 2.16 0.1671
T2 1 106.6056250 106.6056250 2.83 0.1183
T1*T2 1 21.8556250 21.8556250 0.58 0.4609

Output 77.2.1 indicates that the overall model effect is not significant at the level, and Output 77.2.2 indicates that neither treatment is significant at the level.

The following statements invoke the ROBUSTREG procedure with the same model:

proc robustreg data=recover;
   class T1 T2;
   model time = T1 T2 T1*T2 / diagnostics;
   T1_T2: test T1*T2;
   output out=robout r=resid sr=stdres;
run;

Output 77.2.3 shows some basic information about the model and the response variable time.

Output 77.2.3 Model Fitting Information and Summary Statistics
The ROBUSTREG Procedure

Model Information
Data Set WORK.RECOVER
Dependent Variable time
Number of Independent Variables 2
Number of Continuous Independent Variables 0
Number of Class Independent Variables 2
Number of Observations 16
Method M Estimation

Summary Statistics
Variable Q1 Median Q3 Mean Standard
Deviation
MAD
time 25.5000 31.2000 34.7500 30.7688 6.6425 6.8941

The "Parameter Estimates" table in Output 77.2.4 indicates that the main effects of both treatments are significant at the level.

Output 77.2.4 Model Parameter Estimates
Parameter Estimates
Parameter   DF Estimate Standard Error 95% Confidence Limits Chi-Square Pr > ChiSq
Intercept     1 36.7655 2.0489 32.7497 40.7814 321.98 <.0001
T1 0   1 -6.8307 2.8976 -12.5100 -1.1514 5.56 0.0184
T1 1   0 0.0000 . . . . .
T2 0   1 -7.6755 2.8976 -13.3548 -1.9962 7.02 0.0081
T2 1   0 0.0000 . . . . .
T1*T2 0 0 1 -0.2619 4.0979 -8.2936 7.7698 0.00 0.9490
T1*T2 0 1 0 0.0000 . . . . .
T1*T2 1 0 0 0.0000 . . . . .
T1*T2 1 1 0 0.0000 . . . . .
Scale     1 3.5346          

The reason for the difference between the traditional ANOVA and the robust ANOVA is explained by Output 77.2.5, which shows that the fourth observation is an outlier. Further investigation shows that the original value of 24.4 for the fourth observation was recorded incorrectly.

Output 77.2.6 displays the robust test results. The interaction between the two treatments is not significant. Output 77.2.7 displays the robust residuals and standardized robust residuals.

Output 77.2.5 Diagnostics
Diagnostics
Obs Standardized
Robust Residual
Outlier
4 5.7722 *

Output 77.2.6 Test of Significance
Robust Linear Test T1_T2
Test Test Statistic Lambda DF Chi-Square Pr > ChiSq
Rho 0.0041 0.7977 1 0.01 0.9431
Rn2 0.0041   1 0.00 0.9490

Output 77.2.7 ROBUSTREG Output
Obs T1 T2 time resid stdres
1 0 0 20.2 -1.7974 -0.50851
2 0 0 23.9 1.9026 0.53827
3 0 0 21.9 -0.0974 -0.02756
4 0 0 42.4 20.4026 5.77222
5 1 0 27.2 -1.8900 -0.53472
6 1 0 34.0 4.9100 1.38911
7 1 0 27.4 -1.6900 -0.47813
8 1 0 28.5 -0.5900 -0.16693
9 0 1 25.9 -4.0348 -1.14152
10 0 1 34.5 4.5652 1.29156
11 0 1 25.1 -4.8348 -1.36785
12 0 1 34.2 4.2652 1.20668
13 1 1 35.0 -1.7655 -0.49950
14 1 1 33.9 -2.8655 -0.81070
15 1 1 38.3 1.5345 0.43413
16 1 1 39.9 3.1345 0.88679