我试图估计汉密尔顿(1989)的基本马尔可夫转换模型,如电子观点网页中的帖子.该模型本身就是RATS中现有的精确复制.
这是示例的时间序列:
gnp <-
structure(c(2.59316410021381, 2.20217123302681, 0.458275619103479,
0.968743815568942, -0.241307564718414, 0.896474791426144, 2.05393216767198,
1.73353647046698, 0.938712869506845, -0.464778333117193, -0.809834082445603,
-1.39763692441103, -0.398860927649558, 1.1918415768741, 1.4562004729396,
2.1180822079447, 1.08957867423914, 1.32390272784813, 0.87296368144358,
-0.197732729861307, 0.45420214345009, 0.0722187603196887, 1.10303634435563,
0.820974907499614, -0.0579579499110212, 0.584477722838197, -1.56192668045796,
-2.05041027007508, 0.536371845140342, 2.3367684244086, 2.34014568267516,
1.23392627573662, 1.88696478737248, -0.459207909351867, 0.84940472194713,
1.70139850766727, -0.287563102546191, 0.095946277449187, -0.860802907461483,
1.03447124467041, 1.23685943797014, 1.42004498680119, 2.22410642769683,
1.3021017302965, 1.0351769691057, 0.925342521818, -0.165599507925585,
1.3444381723048, 1.37500136316918, 1.73222186043569, 0.716056342342333,
2.21032138350616, 0.853330335823775, 1.00238777849592, 0.427254413549543,
2.14368353713136, 1.4378918561536, 1.5795993028646, 2.27469837381376,
1.95962653201067, 0.2599239932111, 1.01946919515563, 0.490163994319276,
0.563633789161385, 0.595954621290765, 1.43082852218349, 0.562301244017229,
1.15388388887095, 1.68722847001462, 0.774382052478202, -0.0964704476805431,
1.39600141863966, 0.136467982223878, 0.552237133917267, …