Fisher's F-test Research Articles

Inter-observer variability in a computer-assisted optic nerve head assessment system

A computer-assisted optic nerve head analysis system was used to measure parameters of the optic nerve head from one eye from each of 15 subjects. The subjects comprised three groups of five, being glaucomatous, ocular hypertensive, and normal. The 15 images produced were analysed in a randomized masked fashion by ten observers of varying levels of experience to test the inter-observer variability of the measurements. An analysis of variance demonstrated significant differences in the measurements made by the different observers. Fisher's F-test showed that all the parameters assessing the pallor area of the optic nerve head had significantly lower inter-image variance (reported previously) compared with inter-observer variance (at the 5% level). Variability in taking measurements of the pallor area of the optic nerve head is mainly due to observer variations rather than the image variations.

Inter‐ and intra‐image variability in computer‐assisted optic nerve head assessment

A computer-assisted system for optic nerve head analysis was used to measure one eye from each of 15 subjects. These comprised three groups of five, being glaucomatous, ocular hypertensive, and normal subjects. Each eye was imaged with the system 10 times. The 150 images produced were analysed in a randomized masked fashion to test the inter-image variability of the measurements. Linear pallor: disc ratios had mean coefficients of variation of 6.6% (normal), 4.8% (ocular hypertensive) and 2.6% (glaucomatous). This trend of reducing coefficient of variation from normal to ocular hypertensive to glaucomatous was reflected in other parameters also. Intra-image variability was assessed by analysing one image from each subject 10 times, in a randomized masked fashion. The mean coefficients of variation of the linear pallor: disc ratios were 6.3% (normal), 3.6% (ocular hypertensive) and 3.2% (glaucomatous). Fisher's F-test showed that no parameters had significantly lower intra-image rather than inter-image variance (at the 5% level). Variability is apparently due to operator variations rather than the image variations.

Viral integration, fragile sites, and proto-oncogenes in human neoplasia

To evaluate the trend of viral integration in the human genome, chromosomal localization of five DNA-containing viruses compiled from literature data was compared to the location of fragile sites and proto-oncogenes. A total of 35 regionally mapped viral integration sites from tumors and transformed cells were distributed over 19 chromosomes. Of the 35 integration sites 23 (66%) were at the bands of fragile sites, and 7 were one band away (20%). This statistically defines the correlation as highly significant (P = 0.0000183, Fisher's F-test). Five integration sites did not correspond to the location of a fragile site. Thirteen integration sites and proto-oncogenes mapped at the same bands (37%), 6 (17%) were one band apart, and at 16 integration sites (46%) no proto-oncogenes were localized (P = 0.00491). Eighteen viral integration sites, fragile sites, and proto-oncogenes (51%) were localized at the same bands or one band distant. This clustering of viral integration sites, fragile sites, and proto-oncogenes is statistically highly significant (P = 0.0000118), and indicates nonrandom viral integration in the human genome.

Stability of plasma low density lipoprotein with abnormal glycolipid composition from patients with Fabry's disease

Fabry's disease is a glycosphingolipid storage disease associated with premature generalized arteriosclerosis. Plasma low density lipoprotein (d 1.019–1.055 g/ml; LDL) was isolated from 4 healthy male volunteers (LDL-N) and from 4 patients with Fabry's disease (LDL-F), in whom plasma globotriaosylceramide (GbOse 3) levels were 2–4 times above normal. LDL-N was labeled with 131I and LDL-F with 125I by the iodine monochloride method; both were then injected together intravenously into 4 mongrel dogs. The rates at which they were metabolized were determined by measuring plasma radioiodine levels daily for 7–8 days. No significant difference (Fisher's F-test) was observed between LDL-N and LDL-F in terms of the size of the intravascular compartments (LDL-N, 86.3 ± 4.2%; LDL-F, 94.2 ± 7.7%), biological half-lives (LDL-N, 24.1 ± 4.2 h; LDL-F, 23.5 ± 3.7 h), or fractional catabolic rates (LDL-N, 0.849 ± 0.066; LDL-F, 0.796 ± 0.070). The results indicate that significant abnormalities of the neutral glycosphingolipid composition of LDL, such as occur in Fabry's disease, do not affect the metabolism of the lipoprotein apoprotein in dogs. The arteriosclerosis in patients with the disease is probably due to damage to vessel walls occurring as a result of defective GbOse 3 metabolism and accumulation of glycosphingolipid in the tissue, rather than to abnormal LDL metabolism.