Introduction
Statistical research is one of the most popular modes of examining patterns and variable associations. It entails collecting a large amount of numerical data and analyzing it to make meaningful conclusions. Its three most important phases are data collection, analysis, and interpretation. Often, data analysis can only occur when the researcher has correctly arranged the data and identified the appropriate analysis tool based on their studies’ aims and objectives. SPSS, R, and MATLAB are some examples of common statistical analysis software that researchers utilize often. These tools can produce such statistical measures as chi-square tests, t-test, and parametric analyses that illustrate variable relations, variations, and patterns (Shih & Fay, 2017). However, one needs to possess many statistical analysis skills to use these tools and interpret the outcomes correctly. Statistical research remains popular because it helps formulate and test hypotheses and make generalizations about a population based on the behaviors of a sample obtained from it.
In this research, the author identifies three statistical research articles about diabetes that utilize the chi-square test (also called the X2-test). The researcher then reviews these articles and includes information about their aims, objectives, and results where applicable. None of the three articles that the researcher identifies explicitly describes its research question or specific objectives. The authors also fail to present their hypothesis exclusively. However, in all of them, the aim of the study is clear and well represented. In all of them, the chi-square test proves that the leading risk factors of type 2 diabetes mellitus include body age, mass index, hypertension, blood sugar levels, cholesterol levels, and triglycerides. The risk of a person becoming diabetic increases significantly as they age. Some factors such as dyslipidemia and positive family history become important predictors of diabetes if assessed collectively.
Context Development
Zou et al. (2016) conducted a critical analysis of type 2 diabetes mellitus risk factors and their association. Their work focused on China’s Guilin region and involved 6,660 individuals. The authors selected the participants using a cluster random sampling method and sent them a cross-sectional survey to collect data. The researchers then took the participant’s physical measurements and did liver and ultrasound on them. Zou et al. (2016) also conducted a general laboratory investigation on the participants and asked them to respond to structured questionnaire questions. Once Zou et al. (2016) collected all the needed statistical data, they used a classification tree to analyze diabetes Mellitus risk factors and their relationship. The authors then compared the metabolic and clinical characteristics between patients and the control population. In quantitatively analyzing the association between these risk factors, Zou et al. (2016) utilized the non-conditional logistic regression model. They also relied on version 18.0 of the Statistical Package for the Social Sciences (SPSS) software to conduct all analyses.
Regarding categorical variables, Zou et al. (2016) expressed them as percentages and used the x2-test (Chi-square test) to analyze them. Zou et al. (2016) used the statistical research process to identify 338 individuals with type 2 diabetes; 217 were men. The researchers also realized that individuals with type 2 diabetes also had hypertension, high body mass index, and high total cholesterol from examining the decision tree. They also found high levels of high uric acid and large amounts of low-density lipoproteins. They concluded that type 2 diabetes in most patients results from the interaction of more than one factor, with age, BMI, and hypertension being chief among them. Without statistical analysis, Zou et al. (2016) would have had difficulties performing the study or coming up with conclusions. The classification and regression tree (CRT) model utilized in this regard is a non-parametric analysis applicable to potential interactions between categorical or continuous variables. For the parent node and the child nodes, the minimum sample size was 100 and 50, respectively.
Liu et al. (2019) also used statistical methods to predict the risk of diabetes. They constructed a logical regression model, a decision tree model, and a neural network model to analyze type two diabetes’ risk factors. They then compared the three models’ prediction accuracy by calculating the relative operating characteristic (ROC) curve. Liu et al. (2019) then inputted the data obtained from this calculation back into the three statistical models. Their results showed that type 2 diabetes’ prevalence in 4177 subjects not previously diagnosed with the disease was 9.31 percent. They also found that age, triglyceride, hypertension, alcohol consumption, and total cholesterol are the most influential type 2 diabetes mellitus factors. The neural network model’s, logistic regression model’s, and decision tree model’s prediction accuracies were 0.780, 0.711, and 0.698, respectively. The differences in the area under the curve after back-inputting the data (i.e., back probation) for all the three models were statistically significant.
Liu et al. (2019) created a database to help them finish a consistency test using the Epi Data 3.1 software with double-entry data. For general descriptive analysis and chi-square generation, they utilized version 24.0 of the SPSS software. Liu et al. (2019) also created the neural network, logistic regression, and decision tree models using the SPSS software. Out of the 4177 participants, they randomly selected 2924 individuals, representing 70 percent of the total, to provide the training data set. They also randomly selected 125 participants, representing 30 percent of the total, to offer the decision tree model and the logistic regression model a validation data set. 115 (9.18 percent) and 274 (9.37) people with diabetes fell in the decision tree and logistical regression model data sets. For the neural network model, Liu et al. (2019) extracted one-third of the people from the training set and the testing set, with each cohort containing 193 and 81 people with diabetes, respectively. Liu et al. (2019) included the chi-square test in their neural network model.
Lastly, Urrutia et al. (2021) investigated diabetes mellitus and associated risk factor’s incidence in Spain’s Basque Country’s adult population. For them, the chi-square statistic helped scrutinize relationships and the baseline characteristics of the study’s population. The exercise was essentially a reexamination of an adult population after a seven-year follow-up program. Urrutia et al. (2021) compare their study findings with previous research conducted seven years earlier that involved randomly selected 847 individuals aged 18 and above. Notably, the participants in that previous study came from Spain’s Basque Country and answered questions from a structured survey. They then took an oral glucose tolerance test and underwent a physical examination. In the 2021 reassessment, Urrutia et al. (2021) collected the same variables from 517 participants, 43 of whom had diabetes. After doing some statistical examinations, the authors realized that the diabetes’ cumulative incidence was 4.6 percent for seven years, with 6.56 cases per 1000 person-years being the raw incidence rate. Fifty-nine percent of those who had diabetes were undiagnosed, suggesting that the disease’s incidence in the general public might be higher than previously anticipated.
Urrutia et al. (2021) identified age as the most important diabetes marker regarding the leading risk factors of the disease. Specifically, they found that the disease’s likelihood increases by 1 percent per decade for individuals aged 60 years and above. The authors also identified dyslipidemia, insulin resistance, and prediabetes as other major risk factors. Urrutia et al. (2021) also found an association between diabetes and hypertension, body mass index (specifical obesity), positive family history, and low education levels. It is unclear how a lack of education leads to diabetes, but poor nutrition, ignorance, and lack of exercising increase among the illiterate population, possibly contributing to such lifestyle diseases as diabetes. The authors successfully identified the most important risk factors in diabetes using the univariate analysis. Sex- and age-adjusted multivariate analysis revealed to Urrutia et al. (2021) that the predictive value of some factors (such as waist-to-hip ratio, dyslipidemia, and family history) increased when they were assessed collectively. They performed all the statistical analyses using version 4.0.1 of the R software.
In all three types of research, the authors did not explicitly identify their research questions or objectives. Instead, they stated their studies’ aims, described the methodologies used, discussed the results, and made concluding remarks. The authors in all three studies did not create any hypotheses for their respective studies. Instead, they focused on achieving their aims by creating an effective and reliable research methodology. All the articles utilize the chi-square statistic and other statistical tests to prove that indeed body mass index, hypertension, and high blood sugar levels are among the leading diabetes risk factors. The researchers’ statistical methods simplify and make numerical data meaningful and useful by showing correlations, associations, and patterns. Statistical research’s ability to simplify data also makes it an important tool for generalization. Notably, every research aims to originate specific findings and generalize them back to the population from where the researchers obtained the participants.
Statistical Tool Discussion
One of the major roles that statistics plays in research is that it helps researchers determine observation patterns. It is useful in scientific investigations as they provide evidence of trends and relationships between phenomena or variables. Statistics also help authors to compare observations with theoretical predictions. Mathematical models and theories often provide ideal situations regarding relationships and the interaction of variables (Nibrad, 2019). Through the statistical analysis, researchers can see how well actual events resemble those predicted mathematically. Generally, statistical methods also help researchers apply scientific methods in research, assist in hypothesis formulation, aid in experimental design, provide probability information, and test hypotheses. It is also useful in collecting, organizing, and analyzing data and expressing inference uncertainty at any preassigned probability level.
Examples of common statistical tests include chi-square, mean, mode, ANOVA, and t-tests. They are used to identify patterns, associations, relationships, and variations in phenomena (Turhan, 2020). Often, the effectiveness and reliability of statistical analysis depend on the methods utilized in examining the data and the accuracy of the information used. If the researcher fails to define the statistical data well within the analysis software, they may have wrong and unreliable outcomes. Another disadvantage of statistical data is that one needs to possess many technical skills to do it at a professional level. Although various software that could help individuals do statistical analysis exist, one needs knowledge on how to organize the data, upload it to the analysis software, and interpret the outcomes. Notably, without effective interpretation, statistical data are useless. They only become meaningful when skilled individuals interpret them and conclude them. Statistical data may require cleaning sometimes. Therefore, the information it represents is not necessarily the whole truth. It is liable to be miscued and does not depict the entire story. Most importantly, since statistical research aims to make generalizations back to the population, its accuracy depends on the size.
Conclusion
Researchers use different statistical tests to generate or test hypotheses or examine patterns, variations, and associations, among other things. For example, researchers use the t-test inferential statistics to determine if two group’s means differ significantly. The chi-square test is also another popular statistic used to understand how categorical variables are related to each other. It is also expressed as the X2-test, and its null hypothesis is that no relationship exists between a given population’s categorical variables. It is often effectively utilized in the evaluation of tests of independence by comparing patterns. In diabetes, common risk factors associated with the disease include body mass index, age, hypertension, and blood cholesterol levels. The three research articles examined in this study used the chi-square test effectively to show how various diabetes risk factors are associated with each other. Without using the chi-square test and similar statistical tests, it would have been difficult or impossible to establish with some degree of certainty the identified variable’s association.
References
Liu, S., Gao, Y., Shen, Y., Zhang, M., Li, J., & Sun, P. (2019). Application of three statistical models for predicting the risk of diabetes. BMC Endocrine Disorders, 19(1), 1-10.
Shih, J. H., & Fay, M. P. (2017). Pearson’s chi‐square test and rank correlation inferences for clustered data. Biometrics, 73(3), 822-834.
Nibrad, G. M. (2019). The importance of statistical tools in research. International Journal of Research in Social Sciences, 9(11), 45-54.
Turhan, N. S. (2020). Karl Pearson’s Chi-Square Tests. Educational Research and Reviews, 16(9), 575-580.
Urrutia, I., Martín-Nieto, A., Martínez, R., Casanovas-Marsal, J. O., Aguayo, A., Del Olmo, J., Arana, E., Fernandez-Rubio, E, Castaño, L., & Gaztambide, S. (2021). Incidence of diabetes mellitus and associated risk factors in the adult population of the Basque country, Spain. Scientific Reports, 11(1), 1-8.
Zou, D., Ye, Y., Zou, N., & Yu, J. (2016). Analysis of risk factors and their interactions in type 2 diabetes mellitus: A cross‐sectional survey in Guilin, China. Journal of Diabetes Investigation, 8(2), 188-194.