Close

01/16/2020

The papers published in the major periodicals such as Nature, JAMA, etc., are considered, for the systematic analyses, as representative for the whole of science. The data from these papers constitute a large volume of information that is generally of high statistical quality. On the one hand, they are very easy to read; and on the other hand, it is also well documented and easily accessible that they have generated many new findings. Therefore, scientific literature is highly attractive for the quantitative study of association studies. The meta-analysis as such is not based on experimental and observational data. It is based on observational data where the size is the same (quantitative and statistical) and the direction of each association is the same. The meta-analysis can be considered as the sum of the results of several studies, each dealing with a different subject matter. The size of the meta-analysis for each individual study is therefore the extent of the knowledge of all these studies, that is, the amount of association research that has been done in a particular field.

The papers published in the major periodicals such as Nature, JAMA, etc., are considered, for the systematic analyses, as representative for the whole of science. The data from these papers constitute a large volume of information that is generally of high statistical quality. On the one hand, they are very easy to read; and on the other hand, it is also well documented and easily accessible that they have generated many new findings. Therefore, scientific literature is highly attractive for the quantitative study of association studies. The meta-analysis as such is not based on experimental and observational data. It is based on observational data where the size is the same (quantitative and statistical) and the direction of each association is the same. The meta-analysis can be considered as the sum of the results of several studies, each dealing with a different subject matter. The size of the meta-analysis for each individual study is therefore the extent of the knowledge of all these studies, that is, the amount of association research that has been done in a particular field.
What is important for a prospective evaluation like the above? Prospective evaluation is used when a group of subjects is evaluated at random from the population. Prospective evaluation is a better procedure in case of an association study because it allows us to compare subjects in the comparison group against each other and against the reference group in terms of both the direction and size of the associations between the two groups. If the two groups are different (different study designs, mean values for several variables on each group, etc.), the relationship between the three categories or factors is not clear, thus the interpretation of any result as ‘a relationship’ is not clear. There is a difference between the difference between two groups and the relationship of the group as a whole to the general population. Some research methods may not permit such a comparison. However, for these reasons, prospective evaluation is often recommended.
How do the above two examples compare with the meta-analysis? Prospective evaluation allows the comparison of the results of the various studies; and it is quite straightforward. Therefore, the comparison of the meta-analysis and prospective evaluation is not as straightforward as that of the two example papers. Although the differences in the methods, data processing, statistical approaches, and statistical modeling of the two examples cannot be ignored, the difference in statistical methods between the two approaches is not large. In order to understand these differences, it is worthwhile to consider the differences among different statistical methods, which are described well in “A Primer on Statistical Methods”. However, this approach needs to be kept in the context of each method. The differences among different statistical methods depend essentially on how the data, if properly analyzed, can be compared with one another. In the case of observational studies, the statistical methods used include regression analysis and correlational analyses (including correlation coefficients). There are several other statistical methods used in meta-analysis and prospective evaluation as well, such as the methods described in “The Statistical Analysis of Surveys: the Design of Studies and Methods of Analysis”. In these methods, both the direction and size of the associations are taken into account by combining a statistical method with an estimation approach for the data on the basis of several statistical methods and/or other statistical models. If a statistical analysis method can detect a significant effect, then it is used to assess the size of the effect(s). In case of longitudinal association studies, the statistical methods used include logistic regression and hierarchical regression analysis.