Epidemiology is a field that seeks to evaluate the validity of study results by assessing them on the basis of three alternative explanations, namely: confounding evidence, randomness and the degree of bias (Aschengrau & Seage, 2008). This requires the use of P-values which are measurements that ascertain whether the outcome of an experiment may have been as a result of chance. According to Aschengrau & Seage (2008), a P-value is a continuous statistic that ranges form 0.0-1.0.
When this value is low, it indicates a low compatibility degree with the null hypothesis and observed data. This value assumes that the process of the experimental study was exceptional and that there was no real difference between new and old outcomes. It simply provides the affirmation that the results were statistically significant. As such, the obtaining of the p-value is the representation of the probability of coming up with an outcome of at least the same extremity as the one obtained in clinical, or biological experiment, or epidemiological study (Elstein, 2004).
The null hypothesis gives a better explanation of data depending on the stipulation that the null hypothesis is true. When the P-value is higher than the null hypothesis, there is a high compatibility degree with the null hypothesis and observed data (Aschengrau & Seage, 2008). According to Cohen (2000), the null hypothesis cannot be proved, but it can be disapproved. In epidemiology, it is possible to quantify random errors where approximate or exact values can be used by estimating the confidence interval. It is easy to use approximate values to calculate confidence intervals as they assume level distribution (Aschengrau & Seage, 2008).
Calculation for the measure of a treatment affecting the confidence interval shows the range of the possible effect. A confidence interval puts the lower and higher ranges on the capacity of any real effect thus aiding in the interpretation of clinical data (Cohen, 2000).
The use of the confidence interval makes it possible to ascertain the achievement of statistical significance efficiently just like conducting a null hypothesis analysis (Thompson 2002). For instance, if the confidence interval is inclusive of values that represent no effect, then this represents a range that is of statistical insignificance. That, for a 95% confidence interval, is of 5% non-significance level. In the same way, if the significance level is noninclusive of the no effect reflecting value, then the range is of statistical significance. That, for a 95% confidence interval, is of 5% significance level. There can be criticism of the confidence interval when there is a misinterpretation of p-value, such as equating the value with the probability that the p-value is real (Aschengrau & Seage, 2008).
According to Cohen (2000), the sample or population sizes has a directly proportional correlation effect on the confidence interval and p-values. This is to say that the bigger the sample size gets, the larger the positive affect it has on the confidence interval. This, in turn, rates the P-value as highly significant in the interpretation of data that are more distinguishable. This is to mean that it is highly precise in determining the significance or insignificance of data in an experiment. The results prove with limited doubt on the significance or insignificance of the experimental data.
In conclusion, the determining of the significance of an experiment should be provided through the confidence level, or together with the p-value. This is mainly because of the confidence level being more precise due to the inclusion of both the upper and lower ranges on the capacity of any real effect and, therefore, aiding in the interpretation of clinical data with expression of effect of sample size.
References
Aschengrau, A., & Seage, G. (2008). Essentials of Epidemiology for Public Health. Burlington: Jones & Bartlett Publishers.
Cohen, J. (2000). Statistical Power Analysis for the Behavioral Sciences: Null Hypothesis. Hillsdale: Erlbaum publishers.
Elstein, A. (2004). The origins and Development of Evidence-Based Medicine and Medical Decision Making. Chicago: University of Illinois Press.
Thompson, B. (2002). What Future Quantitative Social Science Research Could Look Like: Confidence Intervals for Effect Sizes. Educational Researcher journal, 6(2), 22-26.