Time Series Techniques For Multivariate Regression Models: A Comparative Study

Time series data is common around us and has a significant role in analysing trend. We explore the connection between two different time series: environmental data and hospital admission rates. The relationship between pollution exposure and its response to health impact has gained interest among researchers recently. Multiple studies have explored the correlation involving atmospheric variables such as PM2.5, PM10, NO2, SO2, atmospheric temperature and relative humidity with schizophrenia. The majority of existing literature used basic data analytics techniques like regression, correlations, and distributed lag nonlinear models to analyse the impact of air quality on schizophrenia. Conventional models like ARIMA and VAR might be suitable to analyze this multivariate time series problem and evaluate their performance in the prediction of the daily incidence of schizophrenia based on environmental attributes. Techniques like ARIMA are univariate time series models that incorporate seasonality and autocorrelation in the data. Another distinct type of time series model for the multivariate lag problem is vector autoregression (VAR). VAR is ideal for collecting several related time series with a lag effect. We evaluated these time series techniques to analyze this multivariate time series between hospitalisation and atmospheric conditions and its lagged association. This comparative study found that the effect of pollutants and air quality was statistically significant in schizophrenia admissions in city hospitals, and ARIMA performs better than VAR.