Finding the Best Model in Nonlinear Regression: Using the Coefﬁcient of Determination

In Indonesia, inﬂation plays a signiﬁcant role in shaping economic growth. Therefore, it is essential to examine the impact of inﬂation on economic growth through a comprehensive analysis. This analysis aims to identify the factors inﬂuencing economic growth in Indonesia by utilizing nonlinear regression analysis. The study focuses speciﬁcally on modeling economic growth in Batam City and its correlation with inﬂation. The primary goal is to identify the most effective nonlinear regression model that accurately represents the relationship between economic growth and inﬂation, as determined by the co-efﬁcient of determination. The method used in this research is nonlinear regression methods provide a more accurate and comprehensive analysis when dealing with complex relationships and can help uncover valuable insights that may be missed by simpler linear models. The results of the analysis ﬁnding the model that is suitable for modeling inﬂation on economic growth is a quadratic model with a coefﬁcient of determination of 73.4%. The research has found that the best model for explaining the impact of inﬂation on economic growth is the Quadratic model with an R-value of 0.734 or 75%. These results indicate that the Quadratic model can account for 75% of the inﬂuence of inﬂation on economic growth.


A. INTRODUCTION
During the current COVID-19 pandemic, the economic development of an area is a regional economic development planning process that is carried out in stages by local governments and communities in managing and allocating existing resources in an area such as Natural Resources (SDA), Human Resources.(HR), Investment and Institutional Resources and other resources to stimulate development both in terms of quantity and quality of economic activity in a region (Khoiri et al., 2020).Several factors influencing economic development are PDRB, unemployment rate, IPM and economic growth (Purba et al., 2019).Economic growth is an activity in the economy that has an impact on increasing goods and services produced by the community and the prosperity of the community will increase in the long term from economic development (Yuniarti et al., 2020).There are many factors that are considered to have an effect on economic growth, some of which are inflation and the Consumer Price Index (CPI).Based on data released by Bank Indonesia, the CPI in July 2022 experienced an increase compared to the previous month of 4.35%, mainly originating from administered prices inflation, in the midst of maintained low core inflation and volatile foods which began to decline.Inflation is one of the main concerns for countries in the world today, inflation is at the root of economic dynamics that risks hampering the pace of economic recovery due to the impact of the Covid-19 pandemic, according to the market monitoring system and basic needs, the Ministry of Trade's Consumer Price Index in Indonesia is estimated increased by 23.73%.Inflation that occurred in March 2022 was driven by increasing inflationary pressures in all regions, Sumatra was one of the provinces experiencing the highest inflation (IHK, 2022) as shown in Figure 1.

Figure 1. Inflation Map
Based on research conducted (Widiaty and Nugroho, 2020), the variables that affect economic growth are inflation and foreign debt, inflation that rises within a certain limit can be used as a parameter in assessing an economy.Research conducted (Palindangan and Bakar, 2021), economic growth is not influenced by the unemployment rate.Meanwhile, the research conducted by (Woestho et al., 2021), shows that the regional share and differential shift components have a significant influence on economic growth in certain regions.In addition, poverty reduction can occur due to economic growth and the growth of the financial sector which has increased every year (Ginting and Dewi, 2013).Based on the above description, inflation has a perspective on economic growth in Indonesia.For this reason, a review of inflation will be carried out on economic growth (Widiaty and Nugroho, 2020).The analysis was carried out to determine the factors that influence economic growth in Indonesia using regression analysis, significantly there are several influencing factors such as exports, tax revenues and exchange rates (Yuniarti et al., 2020).In this study, we will utilize nonlinear regression analysis to explore the connection between the economic growth observed in Batam City and the corresponding inflation.Our objective is to identify the most suitable nonlinear regression model that effectively characterizes the relationship between economic growth and inflation, as indicated by the coefficient of determination.
Meta regression analysis reveals that education positively influences economic growth, as evidenced by its significant impact across various influencing factors (Benos and Zotou, 2014).The impact of wealth on long-term economic growth and natural resources has been examined using the Meta Analysis approach, revealing that certain aspects of the study design are particularly effective in explaining the observed variations (Havranek et al., 2016).(Haseeb et al., 2019) Extensive literature exists on supply chain performance (SCP), which is influenced by both regulatory constraints and a company's diverse motivations.Additionally, the analysis indicates that SCP is affected by investment growth rate and population growth rate.On (Kala et al., 2018), Research conducted using the Cobb-Douglas Production Function and the ordinary least square method (OLS) indicates that controlling inflation can lead to a reduction in interest rates, thereby increasing capital accumulation and ultimately fostering economic growth in Indonesia.On the (Yolanda, 2017), Multiple regression analysis was employed to discover various models in the data analysis.The findings, particularly from model 1, suggest that BI Rate, Foreign Exchange Rates, Money Supply, oil price, and gold prices significantly influence the inflation level in Indonesia when considered together.
Conversely previous studies that utilized a single method to model economic growth, this research aims to enhance the analysis by incorporating multiple models and implementing nonlinear regression analysis.The study seeks to explore a range of model options, including linear, logarithmic, quadratic, and exponential models, to determine the most appropriate one for representing economic growth.This comparison of methods enables us to assess and comprehend the extent to which each model reflects the relationship between the variables under study.By doing so, we can identify the model that provides the most accurate and reliable description of the data.Ensuring the model's reliability is crucial in obtaining dependable results.

B. RESEARCH METHOD 1. Descriptive Statistics
Descriptive statistics are part of statistics that play a role in data collection, presentation, and interpretation of data both in graphic and descriptive form, so that the data presented can be easily understood or read (Handayani et al., 2023).Descriptive statistics serves to describe the object under study through sample data or the population that is the object of a research conducted (Haridanti et al., 2018).

Nonlinear Regression Analysis
Regression analysis is a statistical method used to determine the pattern of the relationship between the dependent variable and the independent variable (Arkes, 2019).Based on the pattern of relationships represented, the regression analysis is divided into two, namely, linear regression analysis and nonlinear regression analysis.Nonlinear regression analysis that is often used to model several types of data is the Log-logistic model, the Weibull model, the Gompert model and the logistics model (Pradani et al., 2021).In addition, in nonlinear regression there are several models that can be used such as exponential models, logarithmic models, and quadratic models.The nonlinear relationship of changes in the dependent variable is followed by a variable change in the independent variable within the specified region.In general, the nonlinear regression model can be us equation 1, Information: y i = observation value of dependent variabel.x i = observation value of independent variabel.θ = predicted parameter vector.ε i = error value.
The function f is a regression function that must be estimated and is a nonlinear function in the parameter θ.However, the regression model is the same as the linear regression model except that y i is a nonlinear function of the observed parameters, but in a nonlinear regression model there is at least one derivative of a function that corresponds to one of the observed parameters.To distinguish the difference between linear and nonlinear models, an estimator of the value in nonlinear regression is used.The exponential regression model can be us equation 2, Information: y i = observation value of dependent variabel.x i = observation value of independent variabel.θ 0 , θ 1 = predicted parameter vector.ε i = error value.
The quadratic model is the value of the dependent variable with a linear up or down linear form so as to form a parabolic curve (Utami et al., 2021).The general equation for the quadratic model can be us equation 3, (3) Correlation is a certain value that serves to determine the relationship between the dependent variable and the observed independent variable (Utami et al., 2021).The relationship between variables is said to be strongly correlated when it has a correlation value of −1 < r < 1.The value of r towards -1 indicates a negative relationship between the observed variables.While the value of r towards 1 indicates a positive relationship between the observed variables.
The coefficient of determination (R 2 ) is a value that is able to explain how much (percentage) the relationship can be explained by the independent variable to the dependent variable (Musfiroh et al., 2018).Meanwhile, the mean square error (MSE) is a value that characterizes the level of error in the estimation results for the estimation of the model or forecasting carried out.The smaller the MSE value in an estimation, the results obtained are considered better (Kurniasih et al., 2013).
The data used in this study is secondary data sourced from the (BPS, 2022) report.The secondary data consists of information related to economic growth and inflation in Batam City for the year 2021.The data used in the study was obtained from the (BPS, 2022) report and focuses on economic growth and inflation in Batam City during 2021,  (Jones, 2019) To find out a suitable nonlinear model to determine the relationship that can be explained by the independent variable, namely inflation to the dependent variable, namely economic growth.
The methods used in this research are as follows: 1. Describe the data 2. Identifying correlations between observed variables 3. Perform nonlinear regression analysis with exponential models, logarithmic models, and quadratic models 4. Comparing the value of R 2 and MSE on the modelling results using the exponential model, logarithmic model, and quadratic model.5. Determine the best model to use 6.Interpret the results obtained to determine the relationship between inflation and economic growth that can be represented.The analysis presented in Table 2 reveals that the data observations are related to specific variables.These observations pertain to inflation data, which serves as the independent variable, and economic growth, which is the dependent variable.This data was collected and analyzed over a one-year period to establish their relationships and patterns.

Figure 3. Plot of the relationship of inflation to economic growth
Based on the plot from figure 3, it can be seen that there is a linear relationship between inflation and economic growth.This allows us to carry out further analysis.Furthermore, correlation analysis was conducted to determine the relationship between inflation and economic growth.The following are the results of the analysis carried out using statistical software: Based on the plot above, it can be seen that there is a linear relationship between inflation and economic growth.This allows us to carry out further analysis.Furthermore, correlation analysis was conducted to determine the relationship between inflation and economic growth.Table 3 are the results of the analysis carried out using statistical software: Based on the results of the correlation analysis conducted using the Pearson correlation coefficient, it can be seen that there is a relatively large positive correlation.This can be seen from the correlation value between inflation and economic growth of 0.732.That is, there is a positive relationship that occurs between the observed variables, namely inflation on economic growth.

Nonlinear Regression Analysis
To model the relationship between inflation and economic growth, a nonlinear regression analysis was performed.The primary objective of this analysis was to identify the most appropriate model for describing this relationship.Table 4 contains the results and details of the nonlinear regression analysis conducted to establish the fitting model for inflation and economic growth.The analysis results in Table 4 reveal that the modeling of the relationship between inflation and economic growth involves three nonlinear models.These models include the logarithmic, quadratic, and exponential models, and they are being considered for comparison with a linear model.Based on the analysis, the study explores and contrasts the performance of these various models in describing the relationship between inflation and economic growth.The summary model obtained from Table 5 using a linear model, the correlation results that can be represented by inflation on economic growth are 0.732, which means that the two variables observed are linearly correlated with each other.Meanwhile, the R-Square value is 0.535.This means that 53.5% of inflation is able to explain the relationship to economic growth, the remaining 46.7% is explained by other variables outside of the observed variables.Furthermore, the analysis of variance results in Table 6 provides insights into the impact of inflation on economic growth.This impact is signified by the significance level (sig.)obtained, which is 0.007, falling below the predetermined alpha (α) value of 5% (0.005).The significance of 0.007 suggests a statistically significant effect of inflation on economic growth, as revealed by the analysis of variance in Table 6.Based on the results of the summary model Table 8, using the logarithmic model, the correlation results that can be represented by inflation on economic growth are 0.7, which means that the two variables observed are linearly correlated with each other.Meanwhile, the R-Square value is 0.49.That is, 49% of inflation is able to explain the relationship to economic growth, the remaining 51% is explained by other variables outside of the observed variables.The analysis of variance results in Table 9 highlights the impact of inflation on economic growth.This impact is evident in the significance level (sig.)obtained, which is 0.01, falling below the predefined alpha (α) value of 5% (0.005).The significance level of 0.01 suggests that there is a statistically significant effect of inflation on economic growth, as revealed by the analysis in Table 9.The regression model is presented in Table 10, depicting the relationship between the variables.In the linear model, the regression equation obtained is Y = 115.831+ 80.442 Inflasi.This linear model showcases how the dependent variable Y is influenced by the independent variable Inflasi, as indicated in Table 10.Based on the results of the summary model from table 11 using the quadratic model, the correlation results that can be represented by inflation to economic growth are 0.857, which means that linearly the two variables observed have a greater correlation with each other than the previous model.Meanwhile, the R-Square value is 0.734.This means that 73.4% of inflation is able to explain the relationship to economic growth, the remaining 26.6% is explained by other variables outside of the observed variables.The analysis of variance results in Table 12 demonstrates a significant impact of inflation on economic growth.This influence is evident through the significance level (sig.)obtained, which stands at 0.003, falling below the designated alpha (α) threshold of 5% (0.005).The significance level of 0.003 points to a statistically significant relationship between inflation and economic growth, as indicated by the analysis in Table 12.Based on the results of the summary model from table 14 using the exponential model, the correlation results that can be represented by inflation to economic growth are 0.686, which means that the two variables observed are linearly correlated with each other.Meanwhile, the R-Square value is 0.47.This means that 47% of inflation is able to explain the relationship to economic growth, the remaining 53% is explained by other variables outside of the observed variables.The analysis of variance results in Table 15 indicate a discernible impact of inflation on economic growth.This influence becomes evident through the significance level (sig.)obtained, which stands at 0.014, falling below the predetermined alpha (α) threshold of 5% (0.005).The significance level of 0.014 highlights a statistically significant connection between inflation and economic growth, as demonstrated in the analysis in Table 15.The graphical representation, presented in the plot, serves as a visual tool for understanding economic trends.By offering a visual medium, the plot facilitates a direct comparison between the trajectories of inflation and economic growth.The basis for this comparison lies in the modeling exercises that have been conducted, providing a solid analytical foundation.Through the plotted data, one can easily discern and analyze the relationship between inflation rates and economic growth patterns.The modeling process underpinning this comparison ensures a comprehensive and insightful exploration of the dynamics between inflation and economic growth.

Quadratic Models
The plot provides a graphical representation.It allows for a visual comparison between inflation and economic growth.This comparison is based on the conducted modeling.

Figure
Figure 2. Research Flowchart

Figure 4 .
Figure 4. Comparison plot of the relationship between inflation and economic growth with the modeling

Table 1 .
Table 1 data used: JURNAL VARIAN | e-ISSN: 2581-2017 Economic Growth and Inflation in Batam City in 2021

Table 2 .
Variable Processing Summary

Table 4 .
Model Description

Table 5 .
Model Summary for Linear Models

Table 6 .
Result Analysis of Variance for Linear Models

Table 7 .
Result of Coefficients for Linear ModelsTable 7 displays the regression model, which represents the relationship between the variables.The regression model obtained in the linear form is expressed as Y = 291.868+1.518Inflasi.This model describes how the dependent variable Y is influenced by the independent variable Inflasi in a linear fashion, as shown in Table7.

Table 8 .
Model Summary for Logarithmic Models

Table 9 .
Result Analysis of Variance for Linear Models

Table 10 .
Result of Coefficients for Linear Models

Table 11 .
Model Summary for Quadratic Models

Table 12 .
Result Analysis of Variance for Linear Models

Table 13 .
Result of Coefficients for Linear Models

Table 13
provides the regression model, offering insight into the relationship between variables.In the linear model, the obtained regression equation is expressed as Y = 204.538+4.514−0.01 Inflasi.This linear model illustrates how the dependent variable Y is influenced by the independent variable Inflasi, as revealed in Table13.

Table 14 .
Model Summary for Quadratic Models

Table 15 .
Result Analysis of Variance for Linear Models