Measuring variables, researchable questions

At the beginning of the planning process, it has to be clarified which questions should be answered by the study. In most cases, the research questions are rather qualitative, unspecific statements than precise algebraic formulations. Hence, such statements have to be operationalized into mathematically manageable terms, l.e. measures have to be defined describing the aspects of the system under study that should be investigated. Such measures are called variables. In the context of experimental studies variables have a lot of attributes depending on their purposes. Table 4.9 shows an overview on the most common attributes of variables. 

In both TIMSS and PISA the relationship between home background factors and the various outcome measures eonstitute important parts of the research questions. PISA puts particularly strong emphasis on measuring rich and reliable home background variables since relations between achievement and home background are important indicators for all countries (OECD, 2001). An example will easily demonstrate the strength of this relation. Figure 1 displays how the over-all TIMSS science achievement in one country depends on the number of books at home. Obviously, this variable is an indirect indicator of the socio-economic status of the family, while the books themselves do not have any important direct influence on student achievement. 

The nse of qnantitative and/or qualitative measures is also an important consideration when planning for the data collection phase of critical analysis. Quantitative measnres are typically numerical, and qualitative measures are textual. Qualitative variables are aggregate indicators of the magnitude of concepts, whereas qualitative variables show the character or content of concepts. It is important to keep in mind that neither qnantitative nor qnalitative variables are better or worse. The use of either type of data should be consistent with the research questions that one is asking, and should not be determined becanse one has a preference for statistical or quaUtadve analysis. 

Three kinds of averages are defined from data that have been obtained for research questions. The mean is the point around which the values in the distribution balance it is the mathematical or arithmetic average. In order to calculate a mean, at least internal-level data exist. The median gives information about the value of the middle position in the distribntion. It is the point in the distribution of values at which 50 percent of the scores fall below and 50 percent of the scores fall above. In order to calculate a median, you must have at least an ordinally measured variable. Mode represents the most freqnent value in distribution. The mode is the simplest measure of averages and is, therefore, not viewed as an overly precise or informative measure of average. In addition, the mode is the only measure that is appropriate for nominal data. 

In experimental research, each studied case is generally characterized by the measurement of x (x values) and y (y values). Each chain of x and each chain of y represents a statistical selection because these chains must be extracted from a very large number of possibilities (tvhich can be defined as populations). However, for simplification purposes in the example above (Table 5.2), we have limited the input and output variables to only 5 selections. To begin the analysis, the researcher has to answer to this first question what values must be used for x (and corresponding y) when we start analysing of the identification of the coefficients by a regression function Because the normal equation system (5.9) requires the same number of x and y values, we can observe that the data from Table 5.2 cannot be used as presented for this purpose. To prepare these data for the mentioned scope, we observe that, for each proposed x value (x = 13.5 g/1, x=20 g/1, x = 27 g/1, X = 34 g/1, X = 41 g/1), several measurements are available these values can be summed into one by means of the corresponding mean values. So, for each type of X data, we use a mean value, where, for example, i = 5 for the first case (proposed X = 13.5 g/1), i = 3 for the third case, etc. The same procedure will be applied for y where, for example, i = 4 for the first case, i = 6 for the second case, etc. 

Several projects and consortia were established in order to solve the above-mentioned problems, e.g., the Micro Array Quality Control (MAQC) project, led by US FDA, which main goal is to assess microarray study variability and to develop standards and quality measures for transcriptomics data [50, 51], Another research project, the human embryonic stem cell-derived novel alternative test systems (ESNATS) recently published a paper to address similar questions using human embryonic stem cell-based in vitro test systems for reproductive toxicity by transcriptomics analysis [52], The strong aspect of this study, that it transparently presents difficulties, such as batch effects, and provides analysis strategies including overrepresented transcription factors. It can be used as basis for further development of reproductive toxicity assays based on transcriptomics analysis. 

In any study, it is important that researchers first establish whether or not their data demonstrate a relationship between POP tissue concentration and tissue lipid levels. This is seldom done, as it is typically assumed that such a relationship must exist for lipophilic contaminants. As is compellingly demonstrated by Hebert and Keenley-side47 in their paper To normalize or not to normalize Fat is the question , such assumptions can lead to lipid normalized POP concentrations that are completely at odds with measured wet weight POP values. Further, since factors other than total lipid (such as differences in lipid class, for example) can affect POP levels in organisms, simple ratios (e.g. ng POP/ng lipid) are often inadequate and may actually increase data variability. In many cases, analysis of covariance (ANCOVA) may prove to be a more appropriate method for lipid normalization of POP concentrations47. 

When objective measurement of performance capacities has been incorporated into many clinical trials, concepts and tools from human performance engineering can facilitate the selection of variables and shed some Hght on issues noted above. In either safety- or efficacy-oriented studies, study variable selection can be characterized as a two-step process (1) identification of the factors in question (Table 82.1) and (2) selection of the relevant performance capacities to be measured and associated measurement instruments. This Hnk between these two steps often represents a challenge to researchers for a number of reasons. First, duality in terminology must be overcome. Concerns about an intervention are typically initially identified with negative terms such as dizziness and not in terms of performance capacities such as postural stability. Human performance models based on systems engineering concepts [Kondraske, 1995] can be used to facilitate the translation of both formal and lay terms used to identify adverse effects to relevant performance capacities to be measured, as shown in Table 82.1. 

One of the classic problems encountered in this type of research is that the tools used don t measure the variables in the question asked. Instead, tools are often used that measure convenient variables even though these variables don t address the question. For instance, if you want to know if a particular teaching innovation results in student learning, using a survey that asks students if they like the innovation does not necessarily answer the question of whether the innovation increased student learning. It is true that a positive student perception... 

While clinical trials research incorporates many different components, the focus of this chapter is limited to study questions associated with human performance capacity variables and their measurement... 

---