Statistical Decision-Making in Behavioral Sciences

NB: We do not resell your papers. Upon ordering, we do an original paper exclusively for you.

NB: All your data is kept safe from the public.

Table of Contents

Introduction
Decision Tree
The Treatment-Control Model
First Study of Interest
Second Study of Interest
Application of the Statistical Decision-Making Model
Conclusion
References

Introduction

The behavior of an individual is determined by a number of factors. These factors originate from the people who surround the subject at hand as well as the environment that he/she lives in (Clogg, 2005). By critically analyzing these factors, it is evident that a number of variables play a significant role in affecting behavior of an individual. Examples of these factors include racial background, religion, culture and beliefs, climate, food availability, living standards, and so on. Thus, to ensure that social and behavioral studies are conducted in an effective and efficient manner, several statistical models are developed to ensure that the treatment of variables results in the generation of accurate results. In social and behavioral sciences, statistical models are mainly used to study human behavior within a social environment (Clogg, 2005). Various methods can be applied to these models to generate data that is accurate and reliable. For instance, individuals can be grouped into different categories and specific data collected. Consequently, data can be collected from observation. Once this data has been collected, it is subjected to statistical analysis where inferences are made with respect to the research question and study hypothesis.

Data collected in social and behavioral sciences are mainly analyzed using inferential statistics. In this approach, analyses are conducted to test the hypothesis that a researcher has formulated with respect to the study questions and variables (Lewis, 2003). In this respect, inferential statistics is used to answer questions such as:

Is there any significant difference within the group on a give variable outcome?
Was the difference between the actual and expected value by chance?
Can the results be generalized to represent a larger population?

In this respect, this paper will focus on the application of inferential statistics in behavioral sciences. To achieve this goal, this paper will focus on the decisions that are made in order to select given statistical tests that can be used in a study. Consequently, this paper will focus on a statistical model that has been specifically designed to be used in a particular study, the steps involved, and it application.

Decision Tree

Several considerations are usually put in place in the process of using inferential statistics in a behavioral and social study. Normally, inferential statistics are used to make generalizations of the entire population from a given sample group. In this respect, it is key to ensure that possible errors to the study are eliminated. In inferential statistics, there are two main sources types of errors:

Sampling error
Sampling bias

These errors are detrimental to the accuracy and reliability of the overall results that have been arrived at. Therefore, the overall outcome is usually a difference between the generalized results of a given population with the actual results that were at sample level. This situation makes the sample results not to be considered as the representations of the population from which they were selected from with respect to a specific variable.

Therefore, to ensure that a given statistical model is appropriate for a given study, a researcher needs to calculate the probability value (p value). The p value represents the probability that a given sample is representative of the population that it has been drawn from with respect to the outcome of a specific dependent variable (Lewis, 2003). The hypothesis of a given study should therefore determine whether a difference exists between the sample groups with respect to the outcome of a given variable. The p value should always have a direct relationship with the null hypothesis of a given study. It is due to this fact that researchers consider the p value as a critical measure of decision making since it is used to either accept or to reject the null hypothesis of a given study hence determining whether the null hypothesis is true or false. Furthermore, the p value plays a significant role in determining the probability of arriving at the obtained result in an event where the null hypothesis is true. It is because of this fact that the following treatments usually apply:

If the p value is small, reject the null hypothesis
If the p value is large, accept the null hypothesis

From a critical point of view, the words ‘small’ and ‘large’ are too vague to be useful in decision making. This is because it is difficult to determine how small is ‘small’ and how large is ‘large’. Therefore, researchers can arrive at totally different conclusions based on the same data due to subjective interpretation of data. Normally, while conducting an analysis using p values, researchers have the following questions in mind:

Do we accept the null hypothesis?
Do we reject the null hypothesis?
What is the most convenient p value that should be used as a bench mark?

To assist in decision making, behavioral and social sciences usually consider the p values of 0.05 and 0.01 as their cutoff values (bench mark). In these studies, these values are referred to as significant levels (Clogg, 2005). For instance, when the p value that is arrived at is equal to or less that 0.05, then this value is considered as significant at 0.05 level. The same concept applies when the p value cutoff is set at 0.01 level of significance. Therefore, if the results of a given study are significant at the level of 0.05, this result means that the result that was arrived at will be present within the population at the rate of 5 times out of 100. However, there are instances when researchers reject the null hypothesis while it is true. In statistics, this phenomenon is considered as type I error (Clogg, 2005). However, the probability of committing this error is much higher at the level of significance of 0.05 as compared to the level of 0.01. Despite this fact, many studies still use 0.05 level of significance since they want to determine the true difference that might be present between two different samples.

From the steps that have been presented in this decision tree, it is evident that setting up a hypothesis of a given study is an essential step in behavioral and social sciences studies. After a hypothesis has been set, the next step involves the calculation of descriptive statistics. This is essential as it assists in the determination of the relationship between the various variables of a given study. Consequently, inferential statistics should be conducted to determine the effects of a given variable outcome to human behavior within a social setting. To ensure that the inferences that are made on this study are representative of the actual population, a probability value (p value) is determined and used either to accept or reject the null hypothesis. From the results that will be finally arrived at, one should thus be in a position to draw conclusions based on the research questions and variables of the study.

The Treatment-Control Model

The information from the decision tree is critical in analyzing and interpreting data from a statistical model. As asserted previously on this paper, numerous statistical models have been developed to assist in decision making in behavioral and social sciences. In line with this fact, this paper will propose a new model; the treatment-control model. This model is based on the fundamentals of inferential statistics. Observation is the main method that is employed under this model. These observations are usually made on sample groups after suitable controls have been introduced to the study. In addition to taking a randomized approach to reduce the level of biasness (hence reducing the overall level of error of the study), this model also takes a non-parametric approach that comprises of finite levels that the variables can be subjected to. In this respect therefore, the simplest form through which this model can be applied is by having a sample group as well as a control group. Based on a given population, samples are selected using various criteria into the sample group as well as the control group. Given the fundamentals of this model, samples can be viewed either as:

Observed as being subjected to treatment
Observed as being subjected to control

This thus results in the development of the parameters below:

The average response in an event where the samples were subjected to treatment
The average response if all the samples were subjected to control
The difference between (i) and (ii)

The description of the treatment-control model provides a theoretical framework of its application. However, its actual application might take a completely different approach. For instance, an individual who formed part of the treatment group might decide not to take part in the study. In this respect therefore, such an individual will be treated as being part of the control group. Given its nature, this phenomenon can be referred to as crossover due to the fact that the individual has moved from the treatment category to the control group.

First Study of Interest

The first study that this model might be applied in focuses on the impacts of sports in preventing alcohol and drug abuse among the youth. In this respect, this study will strive to answer the following research questions:

What are the possible causes of alcohol and drug abuse among the youth?
Do sports play a role in reducing the rates of alcoholism and drug abuse among the youth?

From the above research questions, only one study hypothesis can be generated:

Ho: Sports has a direct influence in reducing the rates of alcoholism and drug abuse among the youth.

Ha: Sports does not have a direct influence in reducing the rates of alcoholism and drug abuse.

The variables of this study include:

Independent variable – time spent on playing different sports
Dependent variable – rate of drug and alcohol abuse (per 100 individuals)

Second Study of Interest

The proposed model can also be used to determine human dietary habits and practices. In this respect, this model can be applied on a study that is used to determine the relationship between junk food and obesity in the modern society. From this topic, the following research questions can be generated:

What are the factors that have contributed to the increase of junk food intake in the modern society?
What is the relationship between junk food intake and obesity in the modern society?
What measures can be put in place to ensure that people develop healthy dietary habits?

From the above research questions, only one study hypothesis can be generated:

Ho: Increased junk food consumption results in obesity.

Ha: Increased junk food consumption does not result in obesity.

From the above research questions and hypothesis, the following are the key variables of this study:

Independent variable – Rate of junk food consumption
Dependent variable – Prevalence of obesity in the population

Application of the Statistical Decision-Making Model

The control-treatment model can be applied in the two studies that have been proposed in this paper. By critically analyzing the nature of the variables of these two studies, it is evident that their treatment using this model will be relatively the same. In this respect, the first step of this study will be to determine the method that will be used to select the sample groups for these studies. From practice, random sampling approach has always been considered as the most effective method of selecting subjects and assigning them into various groups of the study. Thus, for each study, there will be a specifically designed treatment group and there will be a control group. It is the treatment group which will be assigned the independent variables in order to have an influence on the dependent variables. After that, the results of both the treatment group and control group will be collected, analyzed, and any difference will be explained using available literature and concepts. Non-parametric tests will be used to ensure that the results of these studies can be generalized to represent the trends within the population with respect to the research hypotheses (whether they will be accepted or rejected).

Conclusion

The treatment-control model can be a very effective tool of inferential statistics that can be beneficial in assisting researchers to make decisions in behavioral and social science studies. By putting different treatments on different sample groups, it is possible to determine the similarities and differences that develop on different subjects hence determining the factors that affect human behavior. Therefore, researchers, managers, politicians, and many other individuals can use this model to develop and implement laws, policies, and strategies that will not only lead to socio-economic development but also result in developments in the understanding of human behavior. At this point however, it is essential to note that despite the fact that the samples in a given experiment might be heterogeneous in nature, no adjustments have been put in place in this model to adjust any statistical differences that might arise as a result of heterogeneity. In the process of generating, preparing, and applying this model, I have come to learn that while generating research questions and hypotheses to be used in a given model, it is critical to determine the relationship that exists between the variables of the study. Consequently, research models are also essential in determining the statistical tests that will be used to analyze the data that has been generated hence affecting the inferences that will be made.

References

Clogg, D. (2005). Philosophical Essays Concerning Human Understanding. London: A. Millar.

Lewis, D. (2003). ‘Causation.’ Journal of Philosophy, 70(1), 556–67.

Click Here To Order Now!