在用负二项回归进行面板数据分析时，想看看高科技和非高科技样本是否有差异，之前可能高科技企业的设定出了些问题，得到的结果不显著，于是我把五家创新性差的企业高科技熟悉从1调整为0，每个样本五年数据，一共也就调了25个数据的高科技属性，结果运行就出现了discontinuous region encountered 有知道是怎么回事的吗
xtnbreg pat21 Hlocation lnSIZE LEV ROA lnage Growth lnreputation nature Vcshare vcgovernance VCnumbers i.year i.Ind if Hitech==0,re

Fitting negative binomial (constant dispersion) model:
Iteration 0:   log likelihood =  -377.6458  
Iteration 1:   log likelihood = -374.13236  
Iteration 2:   log likelihood = -373.47226  
Iteration 3:   log likelihood = -373.31656  
Iteration 4:   log likelihood = -373.28278  
Iteration 5:   log likelihood = -373.27841  
Iteration 6:   log likelihood = -373.27781  
Iteration 7:   log likelihood =  -373.2777  
Iteration 8:   log likelihood = -373.27768  
Iteration 0:   log likelihood = -360.04429  
Iteration 1:   log likelihood = -332.82687  
Iteration 2:   log likelihood = -305.08855  
Iteration 3:   log likelihood = -290.95373  
Iteration 4:   log likelihood = -290.95019  
Iteration 5:   log likelihood = -290.95019  
Iteration 0:   log likelihood = -290.95019  
Iteration 1:   log likelihood = -279.52776  
Iteration 2:   log likelihood = -274.08277  
Iteration 3:   log likelihood = -273.37784  
Iteration 4:   log likelihood = -273.23615  
Iteration 5:   log likelihood = -273.20556  
Iteration 6:   log likelihood = -273.19858  
Iteration 7:   log likelihood = -273.19689  
Iteration 8:   log likelihood = -273.19656  
Iteration 9:   log likelihood = -273.19649  
Iteration 10:  log likelihood = -273.19649  
Fitting full model:
Iteration 0:   log likelihood = -266.19749  (not concave)
Iteration 1:   log likelihood = -263.56032  
Iteration 2:   log likelihood = -258.78153  
Iteration 3:   log likelihood = -257.61796  
Iteration 4:   log likelihood = -256.78707  
Iteration 5:   log likelihood = -256.72948  (not concave)
Iteration 6:   log likelihood = -256.72893  (not concave)
cannot compute an improvement -- discontinuous region encountered
r(430);