Multimodal Distribution over time

Dear All,

I have a time series dataset that looks at discrete events that occur over a specific time period lets say between 1st Jan 2000 - 1st Jan 2010. The events are recorded in serial date format, and the number of repeats corresponds to the number of occurrences (e.g 35431 ,35431, 35431, 35431 = 4 occurrences of 01/01/1997, 36901, 36901 = 2 occurrences of 10/01/2001 and so on).

These events are considered to be normally distributed over that time period. However after looking at a histogram plot I can see that there are 'bumps' at the end of the dataset time period suggesting that the data set is multimodal and that there maybe 3-4 processes that are causing these 'bumps' toward the end of the time period.

So my question is can i separate out the 'bumps' at the end of the datasets statistically into separate regions so i can perform analysis upon then? I have looked at gmdistribution and Principle Component analysis but i can not seem to get them to work.

Thanks Laura

An example of the dataset can be found here:

Column A (Serial Date)

35490 35612 36069 36161 36165 36203 36287 36297 36300 36307 36327 36334 36341 36344 36361 36370 36383 36481 36495 36525 36545 36571 36615 36623 36630 36636 36657 36665 36671 36678 36682 36692 36707 36708 36732 36734 36767 36819 36844 36851 36871 36879 36891 36900 36934 36938 36962 37014 37035 37044 37060 37069 37074 37079 37088 37090 37093 37099 37117 37134 37154 37231 37232 37245 37256 37323 37330 37350 37392 37425 37432 37440 37463 37480 37504 37523 37533 37539 37540 37551 37557 37564 37572 37613 37621 37638 37659 37704 37711 37713 37746 37753 37826 37840 37851 37860 37881 37914 37942 37970 37986 37991 38020 38049 38139 38195 38205 38217 38230 38294 38336 38344 38352 38362 38454 38469 38474 38503 38516 38517 38544 38552 38560 38574 38631 38637 38638 38646 38673 38685 38687 38690 38699 38717 38726 38727 38739 38748 38762 38779 38810 38818 38846 38848 38881 38909 38924 38929 38937 38972 39000 39027 39035 39041 39056 39063 39072 39077 39126 39135 39167 39170 39171 39176 39182 39185 39202 39239 39245 39260 39273 39308 39336 39364 39399 39405 39406 39455 39490 39546 39610 39637 39672 39673 39707 39735 39741 39744 39792 39827 39834 39882 39883 39918 39962 39974 40009 40023 40037 40064 40100 40128 40156 40159 40191 40199 40200 40213 40219 40235 40240 40247 40268

[Tom Lane]

Maybe it's too late for you anyway, but I have trouble understanding what you want. For instance I don't see any repeated values in your example.