Multivariate Time series Modelling

I have a multivariate time series of weather data. It contains 14 variables. I first checked for stationarity using Johansen’s test. All the eigen values returned were <1. Therefore I proceeded with VAR Model for this dataset. I got RMSE values ranging from 8-150 for 13 of the variables, but the prediction for the 14th feature contains ONLY ‘NaN’. What am I missing?
Also, how to interpret if the RMSE values are good enough?
The dataset contains over 4.2Lakh timestamps.
And is there any restriction that Johansen’s test accepts only 12 variables?

© Copyright 2013-2019 Analytics Vidhya