Essentials of Biostatistics. 10. Statistical inference from quantitative data: comparison of means and other locations

Abstract

The focus in this Article is on quantitative measurements that are generally summarized in terms of means. Hemoglobin level (Hb), serum zinc level, blood pressure, anthro-pometry such as weight and head circumference, etc., are examples of such measurements. We discussed in our earlier Articles(1,2) that mean is a statistic that depends on the pattern of distribution of values in the target population. Thus, forms such as Gaussian are specially important to draw inference in this case. Section 10.1 is on comparison of means. This could be with a specified value such as the mean Hb level of children given iron supplementation compared with a prespecified desirable value 14.0 g/dl, or could be the comparison of means of two independent groups such as cmparison of mean birth weight in pregnant women with and without vitamin A supplement. Comparison of means in paired data is given in Section 10.2. Comparing the mean hematocrit levels of a group of children with dengue fever before and after therapy is an example of paired comparison. This section also includes cross-over traials such as measurement of forced expiratory volume (FEV) in asthmatic children receiving two types of treatments in cross-over (or interchange) manner. The methods of these two sections are generally applicable when the underlying distribution is Gaussian. For non-Gaussian forms, particularly when n is small, the methods needed to compare locations are called non-parametric methods. The distribu-tion pattern in some restricted class of subjects such as blood glucose level in the diabetics and triglyceride level in obese can be nonGaussian. Nonparametric methods for comparing two independent and paired samples are discussed in Section 10.3. Section 10.4 is on comparision of means in three or more groups such as mean heart rate of children belonging to families with different dietary patterns. This section describes the popular ANOVA procedure. The last Section is a debate on real meaning of statistical significance so that it is not confused with medical significance.