On 2017 Mar 27, Janet Kern commented:

In the Endres et al. study above, they found a marginally significant trend in decreased glutathione (GSH) signals between the two groups in the dorsolateral prefrontal cortex (p=0.076). It did not quite reach statistical significance. To achieve statistical significance, a study must have sufficient statistical power. Statistical power, or the power of a test to correctly reject the null hypothesis, is affected by the effect size, sample size, alpha significance criterion. In their study, the overall and single-group differences in neurometabolite signals, the level of significance was corrected for multiple tests using the Bonferroni approach (p < 0.025 due to performing the measurements in two independent regions). So, the alpha significance criterion was 0.025. Effect size is the magnitude of the sizes of associations or the sizes of differences. Typically, a small effect size is considered about 0.2; a medium effect size is considered about 0.5; and a large effect size is considered about 0.8. Assuming an alpha, two tailed, set at 0.025, and a large effect size of 0.8 and a power of 0.8, the total number of subjects required would need to be 62 (or 31 in each group). The sample size the Endres et al. study was 24 ASD patients and 18 matched control subjects. Was there truly no statistically significant difference between the two groups, or was the study underpowered?

On 2017 Mar 27, Janet Kern commented:

In the Endres et al. study above, they found a marginally significant trend in decreased glutathione (GSH) signals between the two groups in the dorsolateral prefrontal cortex (p=0.076). It did not quite reach statistical significance. To achieve statistical significance, a study must have sufficient statistical power. Statistical power, or the power of a test to correctly reject the null hypothesis, is affected by the effect size, sample size, alpha significance criterion. In their study, the overall and single-group differences in neurometabolite signals, the level of significance was corrected for multiple tests using the Bonferroni approach (p < 0.025 due to performing the measurements in two independent regions). So, the alpha significance criterion was 0.025. Effect size is the magnitude of the sizes of associations or the sizes of differences. Typically, a small effect size is considered about 0.2; a medium effect size is considered about 0.5; and a large effect size is considered about 0.8. Assuming an alpha, two tailed, set at 0.025, and a large effect size of 0.8 and a power of 0.8, the total number of subjects required would need to be 62 (or 31 in each group). The sample size the Endres et al. study was 24 ASD patients and 18 matched control subjects. Was there truly no statistically significant difference between the two groups, or was the study underpowered?