TABLE 2.

Multiple comparisons and means for groups in homogenous subsets with Scheffé's post hoc tests^a

Group	Mean for subset
	IgG			IgM			IgA
	1	2	3	1	2	3	1	2
DPP IV
Control children	0.1432			0.1694			0.1336
Control adults	0.1994			0.1826			0.1654
Autism		0.3488			0.3260			0.3620
Autoimmune		0.4282			0.3700			0.5120
DPP I
Control children	0.1446			0.1492			0.1508
Control adults	0.1492			0.1480			0.1546
Autism		0.3580			0.2980			0.3320
Autoimmune		0.3940				0.4300		0.4560
CD13
Control children	0.1432			0.1310			0.1484
Control adults	0.1994			0.1400			0.1218
Autism		0.2900		0.1700				0.4040
Autoimmune		0.2460		0.1680				0.3900
Gliadin peptide
Control children	0.1640			0.1460			0.1454
Control adults	0.2140	0.2120		0.1920			0.1860
Autism		0.3600	0.3600		0.3580			0.3680
Autoimmune			0.5060		0.4640			0.4520
HSP-60 peptide
Control children	0.1620			0.1300			0.1618
Control adults	0.1980			0.1770	0.1770		0.1636
Autism	0.2800				0.3380	0.3380		0.6100
Autoimmune		0.5020				0.4880		0.5820

^aMeans for groups in homogenous subsets are displayed based on the type III sum of squares (sample size of 200, with 50 subjects in each group). Means that are reported in the same subset are statistically similar. For example, means for IgG (DPP I) for the control children and control adult groups are 0.1446 and 0.1492, respectively, which are statistically alike. Similarly, the means for the autism (0.3580) and autoimmune (0.3940) groups are statistically the same. However, the means for control children are significantly different from the autism or autoimmune group. Similarly, the means for the control adults are statistically different from both the autism and autoimmune groups. Also note that the control groups (0.1492 and 0.1480) are similar for IgM (DPP I), but autism (0.2980) and autoimmune (0.4300) results are not statistically alike. Potentially, up to four subsets could be formed, simply because for each dependent variable we have four experimental groups. When this is the case, each mean's group should be reported in a separate subset. However, if all four means are statistically alike, all should be reported in one subset. For example, for CD13 IgM, all means are reported in the sample subset, indicating that no difference between the four groups was detected.