How many ways can we show they're early?
 

There are many ways to compare the arrival of Rufous Hummingbirds in 2003 to arrival timing in previous years.  One would be to compare first detection dates for 2003  to first detections dates from past years for the same location.

When we did this we found that some birds, at least, were running ahead of schedule by as much as 4 weeks and averaging about 1 week ahead, when compared to the previous two years.  The arrival of a bird at Ketchikan, AK 2 days ahead of the all time record and about 2 weeks ahead the average certainly continued this trend.

 

 
                        2003    prev            diff
                                earliest
New River,OR            33      40              7
Waldport, OR            33      35              2
Florence, OR            39      41              2
Astoria, OR             39      55              16
Gig Harbor, WA          39      67              28
Nahcotta, WA            42      57              15
Coos Bay, OR            44      47              3
Lyngstad Heights, OR    45      45              0
Nanaimo, BC             48      51              3
Portland, OR            49      55              6
S. Salem, OR            50      66              16
Vancouver, WA           50      62              12
Merlin, OR              51      58              7
Olympia, WA             53      63              10
Castle Rock, WA         53      62              9
Corvallis, OR           54      45              -9
South Beach, OR         56      45              -11
N.Albany, OR            56      76              20
Seaside, OR             56      63              7
McMinneville, OR        56      60              4
Newberg, OR             57      61              4
Montesano, WA           58      67              9
Philomath, OR           58      58              0
                                
average                                         7
Another way to look at earliness is to compare the number of first detections reported on any given day, the freqency of first detection reports.  These can be shown in histogram form. 

The 2003 histogram looks to be shifted forward by about 6 days.  An unpaired t-test produces the following:
 
P value and statistical significance: 
  The two-tailed P value is less than 0.0001
  By conventional criteria, this difference is considered to be extremely statistically significant. 

Confidence interval:
  The mean of Group One minus Group Two equals -6.200
  95% confidence interval of this difference: From -9.071 to -3.329 

Intermediate values used in calculations:
  t = 4.2513
  df = 268
  standard error of difference = 1.458 

What all this means is that we can place a high confidence in the 6 day difference between first detection frequencies for 2003 compared to the the previous two years.  Those of you that are used to looking at graphs will probably note that the 2003 graph could be considered bimodal (it has two peaks).  This second peak comes from reports on the mainland coast of British Columbia and could be an artifact of the cold snap (with snow) that occurred on the dates in between the the two peaks.  But that's just a guess...