Statistics in Research and Analysis

¶ … Statistics in Research and Analysis

The experiments, analysis and statistics-5

Uses of statistics in experiments and research-5

Tools of Analysis-7

Experimental Design-9

Common uses in every day life-12

USE of STATISTICS in RESEARCH and ANALYSIS

This paper concerns itself with the use of statistics as a means and the important tool in research and analysis -- both in the scientific and social sphere. Statistics can be defined as a study of variability and enumeration. It tries to quantify and enumerate uncertain things in a scientific manner. That is because there is an element of uncertainty in all affairs of research and information processing. For example measurements made by scientists and other people are mostly uncertain in some way. One example cited by Kault (2003) is that it is "impossible to know exactly the true average blood pressure of the average healthy person."

Thus when there is a need to conduct a research regarding blood pressure treatment, the variability of the pressure from person to person, and at different times of the day would compound the issue. Statistical formulas are used to generalize the issue so that it can be made universal. It is also necessary in this example to know the effects of the new treatment and a parameter set for the measure of its success or failures. No doubt there would be both failure and success. What is the threshold to be set for the experiment to be declared a success and what must be the threshold for it to be a failure is often determined by the type of data and the type of analysis tools used? All these tools are invariably the subject of statistical analysis. The increased need for information today is general in the sense that researchers are found to tackle many issues in all facets of human existence.

Academic studies in various branches of science have become so specialized that there is a need to process different data in speed and make meaningful information from them. The demand ranges form scientific studies -- physics, medicine and extends to management, finance and the market analysis. It is now necessary for financial analysis. Statistics was a part of the Roman Empire. The purpose of conducting a numerical study is to arrive at the truth of a situation and this means that the study ought to be presented in such a way that on perusal the required essence of the data is visible on examination and especially in a large data set, to make meaning out of the collected information tools have to be used to make it concise and representative of the whole population of the information data.

We can surmise it from the fact that Roman governments conducted annual census of their population. Thus it was a part of governance to get at the data of the people so that policies could be formulated based on their figures. The tools for this and the use of statistics as a tool to arrive at meaningful pointers from other wise jumbled information thus is a part of the general quest of humankind to find the truth and apply it to necessary situations. It was especially in the twentieth century, that scientific method was sought to be used in all spheres of research and in many areas of knowledge, to measure things and then find methods that were error free in the interpretation of measurements with the outcomes.

These bring to the fore the importance of statistics and the method by which it can be used as a tool in research and analysis.

The experiments, analysis and statistics

Statistical analysis is used in cases where the statistical concept of true value -- namely the ability of the analysis to be repeated makes it a choice in scientific and social enquiries. Thus in researches conducted in natural sciences, the problem can be simplified by controlling the conditions, and here statistical methods can be used to design and analyze the results of the experiments making them very reliable and error free. There could also be a link created by statistics between interrelated disciplines where the outcome of one experiment say in biology could also be used in another research as in medicine. However where the science is not exact like in social sciences, there cannot be controlled measurable experiments and these investigators have to rely solely on the statistical means, and statistical analysis is used in "the study of occurrences where the true value or relation cannot be measured."

There is no science that can conduct an experiment without the use of statistical tools to interpret the results.

Uses of statistics in experiments and research

Having noted that statistical tools are the basis of all research and information processing, it is pertinent to explore the role of statistics in research. Statistics form the basis of empirical research. This is because the research is made to answer questions about variables and see if they are related. For example a teacher may want to know if the method of teaching that is being used is more effective than another contemporary method, or a psychologist may want to determine the relation between crime and economic factors. This leads to the process of collecting and analyzing data. This is the basis of statistical operations. Researches cannot be effective without data analysis and statistics is the means of this analysis. The research question can thus be answered only by using statistical methods that either is quantitative, qualitative or a mixture of both.

Researches in statistical models begin with the classification and identification of variables. The oldest explanation of variables is that it is a "characteristic or attribute that varies from item to item in a population."

For example while the larger data set may include 'vehicles running on petroleum gas' some of the variables could be cars, buses, locomotives, airplanes and so on. All these variables are connected to each other by the central description namely the use of petroleum gas. It is defined that the variable divides a population into various groups. Variables can be numerical -- countable like the number of males inside a room.

It can be thus quantified. It can be qualitative in the sense that it can have attributes like 'good', 'efficient' or a historical or observational base that cannot be strictly numerical. Emotional responses for example -- how much one likes ice cream cannot be strictly quantified. Statistics by itself is a science that has many tools that could fit individual research necessities and one set of research need never be identical to another. Both quantitative and qualitative data can be analyzed with appropriate methods while designing experiments and proper choice of the statistical concept that could closely fit the design. Thus all experiments boil down to finding the relation between variables, and finding useful information from them.

Tools of Analysis

There are many tools from the average to multivariate analysis that can be used depending on the need. The basic requirement is to find the variables that would be required to be used in order to find the method of using data. In considering these methodologies the important thing that stood out was that the samples and variables come from two different strata that are opposed in a way. In a case where there are distributed variables with multiple possibilities, researchers use the 'multivariate analysis' especially the partial correlation. Partial Correlation is a system that could be designed to firstly identify the variables has the strongest relationship with the dependent variable when all other variables in the model are controlled. This is the best choice in most focused studies. These methods are ideal for variables that have to have some relation between them or as pointed out in the preceding arguments, have a relation in a spectrum. Some other type of analysis has also been recommended by authors who wrote on business research -- like 'path analysis.'

The "importance of these multivariate techniques lies not in their statistical sophistication but in their value for testing complex theoretical models… These techniques help us evaluate our models and encourage us to specify our model clearly but they do not create models out of thin air."

Thus the multivariate analysis for a correlation coefficience would in truth be enough. In fact the researcher has many models like the 'non linear structural equation model' and the 'Kenny-Judd model with interaction effects' could be useful in business analysis.

Such models are competent where the interaction effects of two latent variables and suggested using product variables to estimate the model. As an example the non-linear equation of this model is as follows: "y =

1 + ? 2 ? 2 + ? 3 ? 1 ? 2 + ?,(1) -with the idea a that there is an interactive effect of ? 1 and ? 2 on y, in addition to the direct effects of ?1 and ? 2 alone, and this interactive effect is manifesting itself in terms of an effect of the product of ?1 and ? 2.."

Where the data consists of numerical things like number of cows that give birth to bulls in a region, the answer may be straightforward. But where there is interaction between the chosen variables, especially where the humans are involved as a variable unlike inanimate objects like gases or salt will not produce the same linear results that could be expected from a scientific experiment as in physics for example. In contrast, the interaction between the multifarious individuals that comprise of the data collected may actually delay or change the patterns of the results based on many factors "that actually may dampen the individual effects of the two variables, as when two noises combine to create a zone of apparent quiet. Two gases may be relatively harmless when released into the atmosphere separately, but may yield lethal toxins when released together."

When we test the interaction effects in the case of the managers and customers of the bank, and try to establish the level of the customer relationship management, there is the same dilemma. Statistics can be used at best when the researcher has designed the experiment properly. For example where human feelings and interactions are involved, individual responses that are prompted by extraneous factors like personal feelings, the different view, or definition of relation and service that an individual may have, it is pertinent at this juncture to see if modern analysis methods that have evolved specifically for business analysis will fit the case. The importance of the method and the tools can be seen if we analyze a real experiment. For example if we were to study the fact that cat owners tend to get diseases from their pets, there are many variables that have to be considered. For example does the owner fondle his cat? What are the precautions that the cat owner takes to keep off from being infected? Here individuals participating vary and are unpredictable variables. Statistics has tools that are effective even in such dilemma. The method used is based on the experimental design.

Experimental Design

In this hypothetical experiment the cats and their owners have to participate and as a control group the individuals who have no feline pets are used. The experiment is designed keeping in mind that in scientific experiments involving humans, the experiment becomes a well planned observational process by which a question can be answered to certainty or an understanding can be reached of the external world. This is done through the observation-hypothesis-experiment. It begins with a chance observation of a new phenomenon.

The important part in the design is finding the appropriate variables. Therefore the experiment has two sets of participants -- one being the households that have cats, and another set in equal number that do not own or have cats. It boils down to a single variable if the family has a cat or not. This is the use of a single variable but not suited to this purpose although the primary position is that it is very easy to summarize results in the case of a single variable. Normally a research cannot be done in the boundary of a single variable but rather the interconnectedness of the variables is the subject of the study. Thus two variables if proved are related, could help in using the information about one to predict the other. Thus in this case the two variable models where the use of one variable is used to predict the probability of the outcome of the other is the bivariate regression model.

Another test that has been considered is the chi-square (x2) distribution which is by far the best for data analyses, and can be used to determine if the variables are dependent or independent. These considerations have prompted the following model for this research: In this case it is to be remembered that there are many pitfalls and things that would not be considered and these may lead to errors. Statistics has no answer to inherent error correction methods if the design is faulty. However statistical methods do have inherent error correction facility. For example in the analysis of the hypothetical cat disease, the discussion can go beyond the suspected diseases that the cat can pass on to any disease that is not yet suspected. This can throw more light on the issue. For example itching if noticed with cat owners but not so with the control group can be a positive indication that the itch may be caused by some dealing with the cat.

Then once this is established there can be further investigation into the issue as a sub-research. There are thus very few variables and the outcome will be based on the explanatory variables used to test the main hypotheses and this must be precise measurements that can be used to accurately measure the outcome, and also later measure the impact of the interventions -- as in this case, the reduction of pathology in cat owners who have taken care of their cats following scientific procedures. In order to design the experiment the major considerations that were taken was the fact that the design can validate the data, that the construct of the experiment is valid and that this experiment can be repeated.

Error creeps in at the sampling time. One of the important tools of statistical method for collecting data is the survey. Survey is relied as the best instrument of data collection. In simple terms the survey "is not just a particular technique of collecting information: questionnaires are widely used but other techniques, such as structured and in-depth interviews, observation, content analysis and so forth, can also be used in survey research."