Vascular Abnormalities, Function
- PDF / 156,846 Bytes
- 12 Pages / 504.57 x 720 pts Page_size
- 16 Downloads / 187 Views
Validity J. Rick Turner Cardiovascular Safety, Quintiles, Durham, NC, USA
Definition The primary objective of an experimental or nonexperimental research study is to obtain a valid estimate of the treatment effect of interest. Validity can be divided into considerations of internal validity and external validity (Rothman, Greenland, & Lash, 2008). Both experimental and nonexperimental studies require consideration to be paid to study design, data acquisition, data management, and analysis. If all of these are of optimum quality and there are no imperfections in the study, the study is deemed valid and the correct result is provided. Any imperfections lead to bias of various types. Internal validity addresses the validity of inferences concerning the source population, and external validity addresses the validity of inferences to the general population, an issue also known as generalizability.
▶ Nonexperimental Designs ▶ Statistical Inference
References and Readings Kleinbaum, D. G., Sullivan, K. M., & Barker, N. D. (2007). A pocket guide to epidemiology. New York: Springer. Rothman, K. J., Greenland, S., & Lash, T. L. (2008). Validity in epidemiologic studies. In K. J. Rothman, S. Greenland, & T. L. Lash (Eds.), Modern epidemiology (3rd ed., pp. 128–147). Philadelphia: Lippincott Williams & Wilkins.
Variability ▶ Variance
Variance J. Rick Turner Cardiovascular Safety, Quintiles, Durham, NC, USA
Synonyms Variability
Cross-References ▶ Bias ▶ Confounding Influence ▶ Experimental Designs ▶ Generalizability
Definition Variance is a sophisticated measure of dispersion that takes into account the position of every data
M.D. Gellman & J.R. Turner (eds.), Encyclopedia of Behavioral Medicine, DOI 10.1007/978-1-4419-1005-9, # Springer Science+Business Media New York 2013
V
2026
point about a central value, typically the mean. It is therefore a measure of the variability within a data set. Imagine the following data set: 6, 8, 10, 12, and 14. The mean is the sum of all the values divided by the total number of values, i.e., 50/5 ¼ 10. How can we find a measure that will capture the totality of the dispersion of the numbers around the mean? An initial thought is to calculate the arithmetic distance each number, or score, lies from the mean, and sum these values. This leads to the following: 4, 2, 0, 2, and 4, the sum of which is 0. That is, the total deviation of the scores from the mean is 0. This is actually true for any such calculation for any data set. The mathematics of calculating the mean ensures that the sum of the deviations of any set of scores from its mean is always zero. So, this strategy does not help convey the degree of dispersion around a central value. However, one extra step takes us to a useful strategy: This involves squaring all of the deviations. Given that a negative number multiplied by another negative number produces a positive value, we now get the following for our original data set: 16, 4, 0, 4, and 16, which sum to 40. This value is known as the sum of squares. If most of the scores
Data Loading...
