Statistical inference for persistent homology applied to simulated fMRI time series data