Descriptive analyses
Descriptive and time series analyses were conducted in R (R Core Team, 2022) using packages plyr (Wickham, 2011), tidyverse (Wickham, 2017), DescTools (Signorell et al., 2019), lubridate (Grolemund & Wickham, 2011), irr (Gamer, Lemon, Gamer, Robinson, & Kendall’s, 2012), and tseries (Trapletti & Hornik, 2016).
Frequency distributions and descriptive statistics were generated to describe sampling effort for JEV sero-surveillance (IgG and IgM antibody testing) by pig age group and district throughout the study period. Frequency distributions and descriptive statistics were generated to explore temporal and spatial patterns in aggregated (monthly and annually) IgM seroprevalence (all age groups combined), and IgG seroprevalence in each age group then in all age groups combined.
Apparent seroprevalence was adjusted to true seroprevalence to account for imperfect test specificity and sensitivity (Rogan & Gladen, 1978) with 95% confidence intervals according to Blaker’s method (Blaker, 2000) in the R package, epi-R (Stevenson et al., 2017). Linear regression was used to quantify the secular trend and seasonality, followed by time series decomposition to visualize temporal components including trend, seasonality (annual cyclical variation), and the remainder, or random, component that could not be accounted for by trend or cyclical variation. A heat plot (function ‘geom_tile’ in R package ggplot2; Wickham (2009)) was generated to visualise the rolling mean of quarterly IgG seroprevalence (the mean IgG seroprevalence across current, prior and subsequent quarters) by district throughout the study period.