Data Analysis for accidents in the Construction Sector (Dissertation Sample)

Contents TOC \o "1-3" \h \z \u Methodology PAGEREF _Toc145092392 \h 2Research Design PAGEREF _Toc145092393 \h 2Data Collection PAGEREF _Toc145092394 \h 2Data Analysis PAGEREF _Toc145092395 \h 2Data Preparation and Cleaning PAGEREF _Toc145092396 \h 3Data Analysis Plan PAGEREF _Toc145092397 \h 3Limitations PAGEREF _Toc145092398 \h 3Findings and Discussions PAGEREF _Toc145092399 \h 4Dominica Social Security data PAGEREF _Toc145092400 \h 4St. Vincent NIS data PAGEREF _Toc145092401 \h 4Exploring Distribution of Injuries in the Dominica Claims Dataset PAGEREF _Toc145092402 \h 4Analysis of Gender Distribution in the SVG Claims Dataset PAGEREF _Toc145092403 \h 5SVG construction safety dataset PAGEREF _Toc145092404 \h 6Heat map of injury vs. cause cross-tabulation PAGEREF _Toc145092405 \h 8Cross Tabulation PAGEREF _Toc145092406 \h 9Grouped bar plot to compare counts of each body part for each injury PAGEREF _Toc145092407 \h 9Heat Map Image (Table (Injury, Body_part)) PAGEREF _Toc145092408 \h 10Discussion and Interpretation PAGEREF _Toc145092409 \h 10Literature Review Insights PAGEREF _Toc145092410 \h 11Analysis of Injury Trends PAGEREF _Toc145092411 \h 12Evaluation of Associated Risk Factors PAGEREF _Toc145092412 \h 13Role of Customized Training Programs PAGEREF _Toc145092413 \h 14Conclusion PAGEREF _Toc145092414 \h 15References PAGEREF _Toc145092415 \h 17Appendix PAGEREF _Toc145092416 \h 18 Methodology Research Design This study will employ a quantitative research design to analyze secondary accident data from multiple sources. Through this approach, trends and patterns in construction accident causation can be examined objectively. A quantitative research design was deemed most appropriate given the retrospective nature of analyzing existing injury records and statistical data. It allows for aggregation of large datasets and identification of relationships between different variables that influence accident causation. The research questions guiding this study pertain to common injury types, affected body parts, root causes, and any relationships between these factors within the construction sector in selected Caribbean nations. A quantitative approach is well-suited to systematically analyze numerical injury records and statistically examine prevalence of different accident characteristics. Quantitative analysis of retrospective secondary data also complements theories and concepts from the literature on accident causation models. Data Collection Accident data will be compiled from three key sources in order to gain insightful sector-specific insights as well as allow cross-validation between datasets. Data will be collected from the following sources for the years 2021-2022: * Injury claims data from Dominica Social Security. This will include details of claim approvals categorized by illness/injury, body part affected, and annual totals. Where available, information on investigations conducted will also be compiled. Records from the Dominica Social Security will provide a national overview of occupational accidents. * Injury statistics from St. Vincent and the Grenadines National Insurance Services. Breakdowns by gender, industry (focusing on construction), and annual numbers will be gathered. * Accident records from a large local construction company operating in St. Vincent and the Grenadines. This dataset promises the most comprehensive information with root causes, body parts, and corrective actions taken for each incident. Where available, data on types of incidents, injuries, affected individuals (gender, age), and timeframe will be compiled. The compilation of detailed accident records from a major local construction firm also promises to yield comprehensive information on each incident such as root causes, impacted body parts, and subsequent corrective actions taken. Data Analysis Upon collection of the secondary data sources, descriptive statistical analysis will be conducted. Frequencies and percentages will be used to examine trends in types of injuries/accidents over time. Frequencies and distributions will be examined for all variables through statistical software to facilitate identification of dominant patterns. Crosstabulation tests will then be employed to investigate associations between incident characteristics where sufficient data allows. Comparison of injury types and affected areas may reveal completeness of reporting to regulatory bodies. Conditional formatting and visualizations will enhance communication of trends. Differences in metrics between time periods and geographical locations will shed light on impact of temporal/regulatory changes. Outliers and anomalies will be flagged for further exploration. The literature review findings will aid interpretation of causation implications from statistical analyses. Comparative analysis between the national statistics and company records will provide insight into completeness of reporting. Insights can also be gleaned by comparing data between the two geographical locations of Dominica and St. Vincent and the Grenadines. Comparative analysis between time periods can indicate if safety practices are improving. Mean and median values may shed light on 'typical' construction accidents. Data Preparation and Cleaning Upon obtaining the datasets, the first step will be to import and combine them in RStudio. The 'readr' package will be used to import the CSV files as data frames. These will then be bound together using the 'rbind()' function into a single master dataframe for analysis. Any issues found will be addressed before analysis begins. When importing the CSV files, I will check that the import was successful by printing out the dimensions and column names of each dataset. This ensures the raw data was read in properly before proceeding. Basic data cleaning and checking will then be conducted. The ‘glimpse ()' function will provide an overview of variables to check for errors or inconsistencies. The restructured data will have unique row identifications assigned for traceability. New variables may also be derived, such as categorizing injury severity. This data wrangling prepares the consolidated dataset for effective analysis and interpretation. Data Analysis Plan Frequencies and distributions will be examined. This will facilitate identification of dominant patterns in injury types, body parts etc. Associations between variables will be investigated using chi-squared tests of independence from the 'stats' package. Visualization of trends and relationships is a key part of the analysis. The 'ggplot2' package will be leveraged to create bar plots, pie charts and line graphs. Time series may also be plotted to indicate safety progress. Examining the structure of the datasets more thoroughly using str() is important to understand the variable types and any factors like missing data. I may extract certain information like: Variable names * Class of each variable (numeric, character, factor etc) * Presence of factors and their levels * Number of complete/incomplete cases For the construction company dataset which has the most detailed records, I will pay close attention to variables like: * Injury/accident description * Potential contributing factors * Body part(s) affected * Severity metrics like days lost from work This initial examination allows me to have a firm grasp of what each dataset contains before cleaning and analysis. It also reveals any obvious inconsistencies or issues needing resolution. In the analysis we may choose to visually inspect random samples of the data using View() to flag errors not captured by str(). Subsetting may be used to extract certain observations for a closer look. Documenting these data understanding steps is important for transparency and to inform the pre-processing required before meaningful insights can be drawn. Please let me know if you need any part of this process elaborated on further. Then descriptive statistics can be generated Limitations Limitations include inability to verify accuracy and completeness of secondary data. Lack of detail on root causes in some data sources may limit causal inferences. One limitation is the self-reported nature of data from one company. Under-reporting is possible due to perceived consequences. However, aggregating multiple sources increases reliability. However, quantitative analysis of available statistics can still indicate high-risk sectors, body parts, and common accident types to inform data-driven prevention strategies. The literature review will further enhance interpretation of trends by connecting findings to existing causation models. Findings and Discussions Dominica Social Security data Total injury claims approved increased from 6 in 2021 to 14 in 2022. Most common were dislocations (3), eye injuries (3), and lower limb injuries (2). No death benefits were paid out in this period, suggesting no fatalities linked to work injuries in construction. Unfortunately investigations into the causes of injuries are not reported, limiting insights. St. Vincent NIS data Total accidents recorded a rise, from 43 (42 male, 1 female) in 2021 to 65 (57 male, 8 female) in 2022. Figures only provide totals by gender and year, devoid of injury or body part details. Exploring Distribution of Injuries in the Dominica Claims Dataset The provided R code and output describes initial exploratory analysis conducted on the "dominica_claims" dataset containing injury information from Dominica. The dataset consists of 2 observations with variables for "Year", "Illness_Injury", and "Location". Glimpsing the data reveals it contains numeric values for years 2022 and 2021, and character variables classifying injuries as either "Dislocation" or "Fracture", located in the "Hand" or "Leg". Tabulating the unique levels of "Illness_Injury" shows one observation for each type. Various techniques were attempted to visualize the distribution of the...