Gibbs Sampling Will Fail in Outlier Problems with Strong Masking
研究了吉布斯采样算法在回归模型异常值检测中的收敛性，发现高杠杆点导致的强掩蔽效应会严重拖慢收敛速度，并通过多个例子说明问题。
This pa¡wr discusses tlJe convergen ce of the Gibbs sampIing algorithm when it is applied to the problem of outli<'r detection in regression 1l1odels.Given any vector of initial conditions, theoretically, tll<' algorit 11m COIlVNg<'S f.o t.\H' true posterior distribution.However, tlw speed of convergellce milY sInw dowll in a. high dimensional parameter space where the parameters are higIJIy correlated.w(~ sllow that tIJe effect of the leverage in regression models makes very difficuIt the convergence of the Gibbs sampling aIgorithm in sets of data with strong masking.The problem is illustrated in severaI examples.