Fundamental principles of statistics

If the null hypothesis is rejected incorrectly, this is a type I (a ) error.

Sensitivity = a high % of true + (and a low % of false -). Using diagnostic tests in parallel sacrifices specificity at the expense of sensitivity. It is used when the greatest penalty would come from missing a true case.{4584}

Specificity = a low % of false + (and a high % of true -). Using one or more diagnostic tests in series increases specificity at the expense of sensitivity, and may be appropriate for conditions in which the penalty of false-positive results is high (i.e. declaring a person to be HIV positive).{4584}

Positive predictive value = (true pos)/(true pos + false pos) x 100{4020}; probability that each of the individuals with positive test results actually has the disease. It is greatly influenced by test specificity.{4584}

Negative predictive value = (true neg)/(true neg + false neg) x 100{4020}; probability that they don't have the disease. It is greatly influenced by test sensitivity.{4584}

This is a common method for determining sample size, and is generally appropriate for relatively simple experiments which are likely to be repeated several times over, with slight differences in treatment. The method utilizes the relationship among the following 6 factors:

Types of epidemiologic study designs

1. Case study: descriptive findings from patients, without reference to a control group; useful in establishing a robust case definition for further studies

2. Cross-sectional (prevalence) study: a snapshot of the source population through sampling at a point in time; used to identify risk factors

3. Cohort study: sample the source population prospectively among exposed and unexposed groups to quantify the risks of exposure

4. Case-control study: sample the source population based on disease status and ascertain information regarding the exposure histories among cases and controls

5. Ecologic study: entire populations comprise the observational unit, and obtain measures of disease incidence or mortality among groups differing in exposure status{4584}

Criteria used to evaluate causation of a disease

1. Strength of the exposure-disease association
2. Consistency of the association in different populations under different circumstances
3. Specificity of the association for the outcome
4. Temporal relationship of the exposure preceding the disease
5. Dose-response relationship of increasing risk with increasing levels of exposure
6. Biological plausibility of the exposure-disease relationship in light of the current body of knowledge
7. Coherence of the relationship without conflict
8. Experimental evidence for the association
9. Analogy with other relationships of similar exposure types and outcomes.

These were proposed by Hill in 1965 and have been modified since then by others.{4584}

Therapeutic Index

The margin of safety of a drug is the dosage range between that producing a lethal effect in 50% of the population (LD50) and that producing the median effective dose (ED50) in 50% of the population. The ratio of LD50/ED50 is called the therapeutic index. {2725}

Top

©1999, Janet Becker Rodgers, DVM, MS

All rights reserved.

Comments? Send an email to rodgers@uky.edu