An Analysis of Recreational and Leisure Areas in Polish Counties with the Use of Geographically Weighted Regression

Round 1

Reviewer 1 Report

This interesting paper is about an analysis of recreational and leisure areas in Polish counties with the use of geographically weighted regression.

The aim of this study was to empirically verify the influence of selected socioeconomic variables on the share of recreational and leisure areas in Polish counties in 2022. The analysis was conducted with the use of the GWR method and local Moran's I statistic. These methods have been validated by other researchers who  reported that local models are more effective than the global OLS regression model in  explaining non-stationary spatial phenomena, including factors that affect the share of  recreational and leisure areas.

So the topic is mainstream, falls into the aims of the journal, and surely meets the interest of the typical readership.

The present findings indicate the share of recreational and leisure areas is much higher in cities than in the surrounding counties. Authors found that in addition to environmental conditions, demand, consumer preferences, and land prices, the distribution of  recreational areas was also influenced by spatial development policies and spatial management. From the methodological point of view, annual cadastral data from the Land and  Building register at county level is a valuable source of data which can be used to analyze  the influence of socioeconomic factors on the share of recreational and leisure areas in Poland.

The methodology it proposed is very practical and logic and the metrics chosen are meaningful. The present study confirmed that local models estimated with the GWR method effectively explain non-stationary spatial phenomena. The manuscript is well written and structured. The argument flows plainly and the whole piece is easy to read. The results were convincing and the discussions were strong. The review and citations have duly reflected the latest discussions on the subject.

Author Response

Dear Reviewer,

We appreciate the time and effort that you have dedicated to the review process. We are satisfied that the article meets your expectations and was received positively. Thank you very much for review.

On behalf of the research team,

Marta Nalej, PhD

Reviewer 2 Report

This is an interesting topic for the paper. However, some important questions remain to meet the publication criteria:

There is no specific justification for selecting only five socioeconomic variables, and a multicollinearity test among those variables was missing in the OLS. I am also unsure if there were local multicollinearity issues in GWR as well.

The other important variables relevant to unequally distributed areas for leisure and recreation were possibly missing. What other socioeconomic variables are available from the data at the county-level measurement? Were only five variables available at the county level from the data? If so, how about integrating other reliable data, such as the Census?

There is no specific rationale for why the author should use GWR followed by OLS. In other words, were there statistically significant spatial dependency issues with the dependent variable or residuals? Although the authors mentioned that decreased AIC values represent a better model, a more specific rationale is needed.

In the same context, it is not clear why the authors used both adaptive and fixed kernel functions for GWR modeling beyond the comparison of AIC values.

According to Nichole and Kim (2022), there are previously published articles that used GWR modeling in the parks, leisure, and recreation context. Please add them to the introduction and discussion sections and check which socioeconomic variables were previously utilized to examine the relationship between socioeconomic status and unequal access to parks and recreational areas at the county level.

As the author mentioned in the discussion section, it is challenging to figure out the cause-and-effect relationship. I agree. However, authors should add additional socioeconomic factors to support the relationship that authors insisted on.

Again, this topic is important for sustainable and regionally equivalent development across Poland, with a focus on residential recreational opportunities. However, the small number of independent variables, the absence of other important socioeconomic variables, and the omission of significant statistical tests for both OLS and GWR should be addressed in the revision.

Moderate editing is needed to improve the readability and clarification of the English language.

Author Response

Dear Reviewer,

Thank you for providing us with the opportunity to submit a revised draft of our manuscript. We appreciate the time and effort that you have dedicated to the review process. We have strived to incorporate changes reflecting the bulk of the suggestions provided by your review.

Here is a point-by-point response to your comments and concerns:

There is no specific justification for selecting only five socioeconomic variables, and a multicollinearity test among those variables was missing in the OLS. I am also unsure if there were local multicollinearity issues in GWR as well.

The socio-economic variables used in the study were selected on the basis of a literature review. Those that occurred most frequently in the context of the impact on recreational and leisure areas were heard. Of course, we agree that this does not cover all the possibilities of using socio-economic variables in examining the impact on recreational and leisure areas. Thank you for drawing attention to the multicollinearity test. The variance inflation factor (VIF) has been supplemented in Table 3 (page 12) and in the text (pages 11–12, rows 308–324). Local multicollinearity issues in the GWR model were also examined and information about the local condition numbers (CN) values was added to the text. CN values do not exceed 10 (page 12, rows 333–335).

The other important variables relevant to unequally distributed areas for leisure and recreation were possibly missing. What other socioeconomic variables are available from the data at the county-level measurement? Were only five variables available at the county level from the data? If so, how about integrating other reliable data, such as the Census?

The variables used in the study were selected based on a literature review. We fully agree that these are not the only variables that can affect the share of recreational and leisure areas in the area of counties in Poland. Of course, these were not the only variables available, but those used in the study were the most common in the literature. According to the reviewer's suggestion, also from the later part of the review, the study was supplemented with additional variables:

unemployment rate,

median age of the population,

share of the population with higher education (page 3, table 1; page 5, table 2).

There is no specific rationale for why the author should use GWR followed by OLS. In other words, were there statistically significant spatial dependency issues with the dependent variable or residuals? Although the authors mentioned that decreased AIC values represent a better model, a more specific rationale is needed.

Based on similar studies and the properties of both models described in the literature, we concluded that the application of GWR should be preceded by Ordinary Least Square (OLS) modelling to identify the global, spatially constant impact of socioeconomic factors on the share of recreational and leisure areas in Polish counties. This information has also been supplemented in the text (page 6, rows 210–214). The use of OLS was the first step in the study, which indicated the existence of a global and spatially constant impact of socioeconomic variables on the phenomenon studied. The second step was to estimate the local influence of demographic, social, and economic factors on the share of recreational and leisure areas in Polish countries, which is why the GWR model was used. Since the GWR model gave better results expressed in a lower Akaike Information Criterion (AIC) (OLS – 1093.7006; GWR – 967.7009) and a higher coefficient of determination R2 (OLS – 0.6974; GWR – 0.8040),and adjusted R2 (OLS – 0.6467; GWR – 0.7673), it was used to examine the influence of socioeconomic factors on the share of recreational and leisure areas in Polish counties.

In the same context, it is not clear why the authors used both adaptive and fixed kernel functions for GWR modeling beyond the comparison of AIC values.

The adaptive kernel GWR model was chosen because of the structure of the data describing the studied phenomenon. The observations are clustered so that the density of the observations varies in the study area. When using an adaptive kernel, the bandwidth is specified as the number of data points in the local sample used to estimate the parameters. Therefore, based on the literature, we decided that it would be better to use a function that can adapt bandwidth distance in relation to variable density; bandwidths are smaller where data are dense and larger when data are sparse, giving better results on the GWR model. An additional explanation has been added in the text (page 12, rows 338–343).

According to Nichole and Kim (2022), there are previously published articles that used GWR modeling in the parks, leisure, and recreation context. Please add them to the introduction and discussion sections and check which socioeconomic variables were previously utilized to examine the relationship between socioeconomic status and unequal access to parks and recreational areas at the county level.

The literature review in the introduction and the discussion were supplemented with selected articles (page 3, rows 100–104; page 22, rows 826–832; page 23, rows 861–864) on recreational and leisure areas listed in the publication by Nicholls and Kim (2022). Also, on their basis, variables like:

unemployment rate,

median age of the population,

share of the population with higher education,

that may affect the share of recreational and leisure areas in Polish counties were added to the study (page 3, table 1; page 5, table 2).

As the author mentioned in the discussion section, it is challenging to figure out the cause-and-effect relationship. I agree. However, authors should add additional socioeconomic factors to support the relationship that authors insisted on.

The study was supplemented with additional variables that influence the share of recreational and leisure areas in Polish counties in 2002. The obtained results were also included in the discussion (pages 22–23).

Moderate editing of English language required

Thank you for this suggestion. We decided that the text needs to undergo a language check.

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

Reviewer 3 Report

1. The text needs to be further improved, for example, the subscripts of the variables in line 186 and 187 are not correct, which is a common problem in the whole paper.

2. Figure 8(b) blue legend is not used, Figure 8(b) red legend is not used, etc. 

3. Some references are not very relevant and can be removed.

4. Many place names are mentioned in the analysis in "3.Results", but they are not labeled accordingly on the figure, which is not easy to read.

The topic is relevant in the field, but only in terms of application. The paper's findings are consistent with the arguments and address the main issues it raises. Some literature citation correlations are not strong. In the analysis of the result, some figures are not readable because of the lack of place names.

There is a small amount of irregular writing.

Author Response

Dear Reviewer,

Thank you for providing us with the opportunity to submit a revised draft of our manuscript. We appreciate the time and effort that you have dedicated to the review process. We have strived to incorporate changes reflecting the bulk of the suggestions provided by your review.

Here is a point-by-point response to your comments and concerns:

1. The text needs to be further improved, for example, the subscripts of the variables in line 186 and 187 are not correct, which is a common problem in the whole paper.

The text has been checked for correctness of notation and references, the notation of variables along with the subscripts has been corrected to correspond to the variables in the equations.

2. Figure 8(b) blue legend is not used, Figure 8(b) red legend is not used, etc.

The map legends in Figures 8-12 have been corrected. Symbols that do not appear on maps have been removed.

3. Some references are not very relevant and can be removed.

The study was preceded by a thorough literature review, which is why the bibliography is so extensive. However, we agree that not all sources provided are equally important for the issue discussed. The references were re-analyzed and 20 references considered less relevant were removed.

Removed references:

4. Many place names are mentioned in the analysis in "3.Results", but they are not labeled accordingly on the figure, which is not easy to read.

Figures 5-12, in section 3 Results, have been corrected. The names of voivodeships and important names of cities, counties and regions mentioned in the text have been added to the maps.

Moderate editing of English language required

Thank you for this suggestion. We decided that the text needs to undergo a language check.

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

Round 2

Reviewer 2 Report

This is a significant improvement, and I appreciate the authors' efforts to incorporate my suggestions into their revision. One remaining comment: Yes, OLS should precede the conduct of GWR. If OLS does not demonstrate the spatial dependency of error terms or a response variable, the use of GWR is supplementary and not required. The Koenker Statistic and Jarque-Bera Statistic need to be addressed to check spatial non-stationarity and dependency alongside the current OLS results.

Author Response

Dear Reviewer,

We would like to thank for careful and thorough reading of this manuscript and for the thoughtful comments and constructive suggestions, which help to improve the quality of this paper. We have incorporated  the changes as suggested.

Here is a point-by-point response to your comments and concerns:

One remaining comment: Yes, OLS should precede the conduct of GWR. If OLS does not demonstrate the spatial dependency of error terms or a response variable, the use of GWR is supplementary and not required. The Koenker Statistic and Jarque-Bera Statistic need to be addressed to check spatial non-stationarity and dependency alongside the current OLS results.

Information about Koenker Statistic and Jarque-Bera Statistic, their statistical significance, has been added in the context of justifying the use of the GWR model (page 12, rows 338–342).

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

Reviewer 3 Report

The correlation analysis section also allows for a more in-depth analysis of what exactly is causing the positive or negative correlation.

The author has completely revised the obvious textual errors

Author Response

Dear Reviewer,

We would like to thank for careful and thorough reading of this manuscript and for the thoughtful comments and constructive suggestions, which help to improve the quality of this paper. We have incorporated all the changes as suggested.

Here is a point-by-point response to your comments and concerns:

The correlation analysis section also allows for a more in-depth analysis of what exactly is causing the positive or negative correlation.

The Results section has been supplemented with explanations regarding the impact of some variables, such as: Median age of the population (page 19, rows 756–761), Income per capita (page 16, rows 631–633), Unemployment rate (page 18, rows 741–743). The authors believe that the influence of selected variables can be better explained for different types of recreational and leisure areas and requires further analyses, which was expressed in the discussion.

Minor editing of English language required.

Thank you for this suggestion. We decided that the final version of the text will undergo a language check.

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD