I am trying to fit ARIMA(p,d,q) model to chronological accidental death data for the period of 50 years in R software. The data set consists of upward trend clearly. So, to achieve stationarity, I did fist difference of the data and plot acf and pacf of the difference data. But there is no significant correlation has been observed in ACF and PACF except one. But while I did stationarity test by ADF test of the differenced data, the result revealed that the differenced data is still non stationarity. My question is that do I go for 2nd difference of the data?
But auto.arima() function in R selects ARIMA(1,1,0) with drift model.