canberrabirds

Pied Currawong: the facts on abundance trends from GBS data

To: <>
Subject: Pied Currawong: the facts on abundance trends from GBS data
From: "Philip Veerman" <>
Date: Sun, 23 Mar 2008 17:27:47 +1100
Further on this: Martin wrote:
 
"The main thing that can be said about the long term trend for Pied Currawong abundance in the GBS results is that there is no trend.  The values of the correlation coefficients for linear and polynomial trends for 21 and 26 year results for the species range from just over 0.1 to nearly 0.3.  These are well below the values that would indicate a trend significantly different to zero.
 
The facts are that the trends are obvious and strong. There are two trends occurring and they need to be teased apart and seen separately, especially because the biological impacts are different. We should be more concerned about the biological impacts than the statistics. It is during the summer that the PC is a major predator of other birds' nesting and during the winter that the PC spreads seeds of woody weeds. We need to see the GBS data in context. There are plenty on long standing bird surveys around the world that are not continuous and only comprise summer or winter counts. If we did that, then the results would be very different.
 
I have just done a regression analysis for the first 21 years of GBS data. Bearing in mind that the analysis of the GBS does not exist beyond the first 21 years. Although I discounted the odd months out of year 1 and 21 because of the split, in that the survey started in July and ended in June). Winter includes June, July and August and Summer includes December, January and February. Therefore each analysis includes 63 data points. The analysis of variance from the regression is also presented (below).
 
Just imagine that we did a GBS in only summer or winter. Two separate regression analyses for these data show: for winter, a probability value (significance of the ANOVA regression F test) of 2.0769E-05 in other words, the chance that the observed decrease is due to chance factors is 0.000020769. Likewise for summer a probability value of 3.59055 E-14 in other words, the chance that the observed increase is due to chance factors is 0.0000000000000359055. Normally biologists accept as reasonable evidence, anything less than 0.05 (meaning 95% probability). These two trends are way off the scale for any suggestion of no trend! Yet Martin says no trends, just on the simple basis of missing that the two trends are running in opposite ways (an increase and a decrease). The possibility of damage done to the GBS and COG's data by such incompetent analysis and comment is of great concern.
 
It is simply the fact that summer abundance has increased to a similar margin as winter numbers have decreased that appear to confuse those not clear in the biological meaning of the results or capacity of the survey). It was not by accident that I chose this species as one of the three to present the full graph of abundance by month throughout the 21 years of the GBS (on page 90).
 

SUMMARY OUTPUT

Of regression analysis of Pied Currawong Winter values of A

Regression Statistics

Multiple R

0.508654885

R Square

0.258729792

Adjusted R Square

0.246577822

Standard Error

1.426770386

Observations

63

ANOVA

df

SS

MS

F

Significance F

Regression

1

43.34189562

43.3419

21.29118

2.0769E-05

Residual

61

124.1760978

2.035674

Total

62

167.5179934

Coefficients

Standard Error

t Stat

P-value

Lower 95%

Upper 95%

Lower 95.0%

Upper 95.0%

Intercept

7.934359474

0.363835294

21.80756

1.71E-30

7.206825636

8.661893313

7.206825636

8.661893313

X Variable 1

-0.04561298

0.009885272

-4.61424

2.08E-05

-0.06537981

-0.025846155

-0.06537981

-0.025846155

SUMMARY OUTPUT

Of regression analysis of Pied Currawong Summer values of A

Regression Statistics

Multiple R

0.782564696

R Square

0.612407504

Adjusted R Square

0.606053528

Standard Error

0.337899565

Observations

63

ANOVA

df

SS

MS

F

Significance F

Regression

1

11.00449822

11.0045

96.38179

3.59055E-14

Residual

61

6.964743096

0.114176

Total

62

17.96924131

Coefficients

Standard Error

t Stat

P-value

Lower 95%

Upper 95%

Lower 95.0%

Upper 95.0%

Intercept

1.479013121

0.086166484

17.1646

4.95E-25

1.306712539

1.651313703

1.306712539

1.651313703

X Variable 1

0.022983684

0.002341112

9.817423

3.59E-14

0.01830234

0.027665027

0.01830234

0.027665027

 
Also Martin wrote:
I have attached a graph showing A values by week (months are difficult
since the weeks don't all start and finish at the bginning/end of
months) for 21 years and later years.  This clearly shows a flattening
of the seasonal distribution in the recent past with more birds in the
warmer months and less in the colder months.  There has been previous
correspondence on this list (last year some time) about the reduced
frequency of the huge flocks seen (particularly in the Weston area) in
the early years of the Survey.
 
As for "months are difficult since the weeks don't all start and finish at the beginning/end of months". What a non issue! Weeks and months are arbitrary human constructs and dividing lines. (Indeed none of the weeks and months start on the same day apart from on 1 January but it doesn't matter one iota). In contrast days and seasons are real. However collecting data by week happens to be convenient to our brains, our lifestyles and constraints of designing and printing a chart. The arrangement of the GBS Calendar was a brilliant way to approach the problem, invented by Henry Nix and it and the connection of weeks and months are fully explained in the GBS Report.
 
Looking at the graph presented by weeks tells us almost nothing more that looking at the graph presented by months. It simply provides more data points. Each with a smaller sample size, thus increasing the importance of random sample error for the rarer species. Thus weeks makes it harder to provide the information in a graph and certainly much harder to describe in narrative. Believe me, having written all the text I did for the GBS Report it would have been much more difficult to describe every change in terms of week numbers rather than months, because everyone understands months. 
 
Lastly, not only has there "been previous correspondence on this list ... about the reduced frequency of the huge flocks seen ... in the early years of the Survey. True but I point out that this has been fully described in the GBS Report and the results Martin has now provided since Year 21 are simply a continuation of the same trend that I described therein (which is hardly surprising).
 
Philip
<Prev in Thread] Current Thread [Next in Thread>
Admin

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the Canberra Ornithologists Group mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the list contact David McDonald, list manager, phone (02) 6231 8904 or email . If you can not contact David McDonald e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU