GEOG/ES&P 330
PigeonWatch Hypotheses for Group Projects
GROUP TOPICS:
Group 1 (Ms. Agtang and Messrs. Alvarez, Amer, and Amoyen): Fall 2020 data
Group 2 (Mesdames Angel, Berhe, Gatch, and Goodsell): 2000-2018 dataThis project will analyze the Fall 2020 field data and the Fall 2000 through Spring 2018 field data for possible signals of natural selection in the dstribution of grouped pigeon morphs by the kinds of habitats in which they were observed. Pigeons are subject to predation by cats and various raptor species. Pigeons feed in social flocks and, when a hawk or falcon goes after them, they scatter in all directions. The raptor will try to focus on one distinctive pigeon in all the confusion. Anything that makes an individual pigeon stand out, like odd coloring, will create a disadvantage and, so, pigeons may be subject to what's called "stabilizing selection" (look that up). Stabilizing selection penalizes morphs that diverge from a common pattern, enforcing greater uniformity in prey species. Think about how all mourning doves and English sparrows look pretty much alike. Feral pigeons, however, are very diverse because they descend from domestic flocks that have been selected by humans to have all kinds of crazy colors and patterns. Why doesn't stabilizing selection drive them back to the wild morph (the blue-bars)? Maybe it's because stabilizing selection operates differently in different environments and, so, selects for different morphs in different places, exploiting that genetic diversity brought in from domestic flocks. Perhaps different backgrounds make one morph less conspicuous than another. Pigeons roam over different locations, though, exposing themselves to different risks. Since they do tend to hang out mostly in about a one kilometer radius, maybe stabilizing selection is "tuning" them to those territories, so that we might see somewhat different mixes of morphs in different kinds of landscapes.
To get at this, you can model your analysis loosely on what you did in Lab 7 by cross-tabbing morphs into different general habitats. Instead of using the two habitats in Lab 7, which you also see cross-tabbed in the PigeonWatchF20.ods and PigeonWatch00-18.ods databases, I'd like you to work with the cross-tabs provided below those. The birds are cross-tabbed into three habitats:
Do a Chi-square with all 7 morphs by the 3 habitats (the new Chi-square plus spreadsheet can handle this). Is there a significant difference in the distribution of morphs by these three core habitats? If so, try to figure out which morphs are over- and under-represented (or roughly proportionately represented) in the three habitats. Create pie charts or bar charts to show the comparison and contrast. Maybe there is something about these environments that make different morphs more survivable in different places.
- urban (including downtowns, industrial, and commercial), habitats with a lot of light grey rooves, sidewalks, and aging asphalt and a lot of humans going about their workaday rounds and leaving garbage
- residential (including suburban and urban residential), habitats with more trees and lawns, a great diversity of roof materials (tile rooves, dark comp rooves, white rock rooves, etc.), fewer humans and their messes
- more natural areas (including beaches and parks), habitats with open spaces, greenery, lake or ocean water surfaces, sand, and often a lot of people and their messes and some people who actually feed birds
Group 1 should do this analysis on the F/2020 data that are pre-processed (cross-tabbed) inBoth Group 1 and Group 2 will use the expanded version of the Chi-square calculation spreadsheet:
https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatchF20.ods starting in the Summary tab (rightmost tab in spreadsheet) cell D21.Group 2 should do this analysis on the F/2000 through S/2018 data that are pre-processed in
https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatch00-18.ods, also in cell D21 in the Summary tab at the far right end of allllll those tabs.
https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/ChiSquareModels5plus.ods, where you can find the 3 x 7 tab. In interpeting your tests, note whether the null hypothesis could be thrown out at the 0.05 standard, how strong the effect size is, and whether we have enough power to trust a finding of no significant difference if that happens. Also, be sure to check the aqua box in the lower left of the Chi-square calculation sheet to make sure that there are no problems with too-small expected cell counts.Also, if the results are significant, try to figure out the pattern of allocation of morphs to habitats. For each morph, is it much more common or much less common than expectation in a particular habitat; others may be somewhat more common or less common in particular habitats; and others may be pretty close to expectation. A color-coded table could do the job, either a second copy of your Chi-square table but with the cells' backgrounds given different colors for much more, more, roughly equal, less, much less common. Or you could create a new table filled with plusses and minuses of various font sizes.
Each group will also prepare three comparison graphics showing the proportion of the seven morphs among the urbanized, residential, and "natural" landscapes. Either three pie charts or one stacked bar chart (where the three stacks are the same height for 100%) would work. I would like the two groups to confer, however, and make sure you both do the same graphs, to make comparison easier in class. Both groups will present on the same day, and discussion can consider whether the two sets of graphs show the same general pattern or diverge through time.
Group 3 (Mr. Gross and Mesdames Gutierrez, Hernandez, and Inthavong): Fall 2020 data
Group 4 (Ms. Kubba and Messrs. Kemp, Kim, and Le): 2000-2018 data
Sexual selection is another theme. Do male pigeons "cruise" potential hens who are similar to them or different from themselves? That is, can we see signs of "assortative mating" among pigeons? Assortative mating occurs when organisms mate with individuals who resemble themselves more than the rest of their species, so that reproductive isolation among lineages can eventually develop even as far as new species. There may be some limitations to this in particular species, if the in group is small enough for inbreeding errors to accumulate (homozygous exposure of deadly recessive genes, which is one reason all human societies have some kind of incest restriction). If there is a tendency to assortative mating, however, it might explain why feral pigeons show such sustained morphic diversity around the world. Maybe this is what is countering stabilizing selection back toward the wild blue-bar morph. This is a key argument that LaBranche makes in the 1999 PigeonWatch backgrounder on the course home page. So, let's see if our data show a significant tendency towards assortative mating.
Group 3 should do this analysis using the F/2020 site database:The PigeonWatchbySite databases have a table way down at the bottom (around K 370) that counts up courtships across all 49 possible combinations of the 7 morphs. Since courting is less common than just seeing the birds, we won't have the statistical power to do a 7 x 7 Chi-square. But we can combine similar morphs and see if there is some kind of sexual selection going on that could account for pigeon diversity. To do this, break out the morphs as follows:
https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatchbySiteF20.odsGroup 4 should do this analysis using the F/2000 through S/2018 site database:
https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatchDatabase00-18.ods
- wild-type (blue-bar)
- melanic (spreads and checkers)
- odd birds (red-bars and reds and albinics or pieds and whites)
This pre-processing has already been done, starting in cell K410. Enter these counts into the 3 x 3 tab of either Chi-square calculation spreadsheet (the original one probably works faster because it has fewer tabs: https://home.csulb.edu/~rodrigue/geog200/ChiSquareModels5.ods). In your analysis, be sure to cover whether the aqua check box in the lower left of the Chi-square sheet has identified any problems with too-small expected values. Whether there are or not, report whether the null hypothesis of no assortative mating can be rejected at the 0.05 level, how strong the effect size is, and whether there is adequate power to pick out a significant association if it's there.
Now, do a companion analysis, where the birds simply prefer similar or different targets than themselves. This entails counting all the times each morph has been "caught" flirting with a target similar to himself and all the times he's been observed hitting up on targets different from himself. The pre-processed counts you need are also in that sites spreadsheet, starting around cell K424. Enter those counts in the Chi-square calculation spreadsheet, this time in the 2 x 3 tab. Again, report on any too-small expected count problems, the fate of the null hypothesis, the effect size, and the power.
Illustrate your results with the Chi-square table and graphics (either three pie charts showing for each male morph who the favored target morphs were or one stacked bar chart showing the percentages of target morphs favored by the three male morphs. Also, do the same with the same-different analysis: three pie charts or one stacked bar chart with three columns just showing percentage same and percentage different. Each group should confer with the other one about graphing choices, just so it's easier to compare your results in class, since you'll be presenting on the same day.
So, what do you suppose is going on with the assortative mating hypothesis? Was Cornell University's Ornithology Lab onto something explaining the sustained morphic diversity of feral pigeon flocks? Or not? Or what?
Group 5 (Mesdames Lopez and Ochi and Messrs. Nguyen and Pham): Fall 2020 data
Group 6 (Messrs. Rosenthal, Ruiz, Schultheis, and Solorzano): 2000-2018 dataLooking at sexual selection another way, are the sexually active birds, whether courting males or courted females, random samples of the regional "bird herd"? That is, do male pigeons considered all together bestow their attentions equitably on any hen candidate they may spot, so that the distribution of their targets resembles the distributions of morphs in the regional populations? And, do target pigeons (presumably females) considered all together get hit on by males drawn randomly from the regional population? In other words, are there "friskier" male morphs and less frisky male morphs? Are there universally regarded "hot" female morphs that all the guys cruise disproportionately?
This time, since we're comparing frequencies of courting males and courted targets against the prevailing proportions of pigeon morphs in the regional "bird herd," we really should use a different kind of Chi-square test, instead of the cross-tab contingency table one you're all so familiar with now (Chi-square test for independence). The better test is the Chi-square goodness-of-fit test, which compares frequencies against known probability distributions, such as the uniform and the binomial (which, of course, you all remember from GEOG 200 or BIOL 260). Alternatively, you can generate expected proportions from a population's parameters. We can consider your 95 site visits and their 1,241 actually identified pigeons as our reference population (Group 5) or the 360 site visits and their 5,796 identified pigeons as the reference population (Group 6). That is what we'll (you'll) do here. To do the Chi square goodness-of-fit test, go to http://www.vassarstats.net/csfit.html.
Group 5 should do this analysis using the F/2020 site database:The information you need is on Sheet 1 of those databases, starting at cell K372. Row 372 gives you the numbers of birds identified in each morph by you and your colleagues in Fall 2020 or by your predecessors from Fall 2000 through Spring 2018. The row of percentages in row 373 includes the unknown and the "other" category, which observers sometimes have to use if the birds took off too soon to be identified and counted. The percentages in row 373 are calculated only for those birds that were successfully identified. Row 375 is our reference probability distribution for known morphs. These seven proportions (blown out to a lot of decimal places to keep down rounding errors) are what you'll copy (one at a tedious time) and put in the VassarStats Expected Proportion column.
https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatchbySiteF20.odsGroup 6 should do this analysis using the F/2000 through S/2018 site database:
https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatchDatabase00-18.odsFor your first analysis, enter each female morph actual numbers (not percentages) from row 391 in the Observed Frequency column. It's probably safest to copy them from the spreadsheet and paste into the VassarStats calculator, just to head off typos. When both the Expected Frequency and the Observed Proportion column are filled out with the seven female morphs and the proportion of each morph from the regional population. Then, click the Calculate button and voilà!: Your Chi-square results. The results are not as detailed as the ones from my Chi-square calculator, so you can't report on effect size or power, but half a loaf, etc.! Once you have your results, be sure to do a screenshot!!! You don't want to redo this tedium! Depending on your operating system, you can use the Microsoft Snipping Tool that allows you to crop before saving (and remember to give it a name and location you can find!). Another way on PCs is to hit the Print Screen button on the upper right of your keyboard next to the function keys. This copies everything on your screen to the Clipboard and then you can open Paint (or GIMP or Photoshop, but Paint will work fine) and paste it in there and then crop it to the VassarStats page and then, again, remember to Save As something you'll be able to find later. With Macs you can get Snipping Tool functionality by pushing Command -- Shift -- 4 and that brings up crosshairs you can use to bracket what you want and then, when you let go of the mouse, it'll save your image as a .png file. You may need to sleuth to find it and then rename it and put it somewhere you can find later.
Now, do the same thing for the male morphs. You can carefully delete each female entry (and keep from having to re-enter all those proportions) and then replace it with the male morphs frequencies summed down the column starting from cell y378. Hit Calculate and there you have it. Remember to screenshot and save it as above.
So, do you have significant results (p<0.05) for either or both genders? You can figure out which morph is courted above her prevalence in the region from the results column called Percentage Deviation and you can tell if a particular morph is significantly elevated in male esteem by looking at the Standardized Residuals. Figure that, for the 0.05 alpha we're using by professorial fiat, that standardized score should be more extreme than + 1.96. Turning to the guys, is(are) there some particularly "forward" male morph(s), morph(s) with high percentage deviations and standardized residuals ≥ 1.96? Might there be significantly less passionate male morph(s) out there, courting "below their weight"?
You should illustrate your results with three pie charts or, alternatively, one stacked bar chart with three columns of percentages. One should be for the regional pigeon population morph distribution, one should be for the target (presumably female) morph distribution and one for the courting male morph distribution. Also helpful for visualizing your results might be tables for each gender showing their percentage deviations and/or standardized residuals with the cells color-coded for significantly over- and under-represented (maybe hot and cool colors?), perhaps a darker shade for those significant at the 0.05 (standardized scores more extreme than 1.96) and a lighter shade for those significant at the 0.10 level (standardized residuals more extreme than 1.58). It would be helpful if you conferred with the other group about your graphing decisions so that in-class discussion is easier.
Group 7 (Messrs. Tejada and Valdez and Mesdames Witke and Zammataro): both Fall 2020 and 2000-2018 data
This project analyzes changes in flock sizes and color morphs over time, both for student data collection throughout the region this semester and for the previous two decades. This will be followed by a more granular analysis of trends in time for four Zip Code Prefixes (the first three digits in a Zip Code are regional complexes of named towns and districts within our vast urban area). I decided that doing this by Zip Code would be TOO granular to see trends (students have observed pigeons in many dozens of Zip Codes over the years and Zip Code level analysis gets us into too much of a small sample effect issue with wild fluctuations in numbers -- and I didn't want this project to be more overwhelming than the others!). So, I identified four Zip Prefixes that have been visited the most by you and by your predecessors: 900 (in and around Downtown L.A.), 907 (taking in Lakewood down to Signal Hill and out to Cerritos, Seal Beach, and Avalon), 908 (Our Fair City), and 926 (including mostly coastal areas from Bolsa Chica down to Laguna Beach and up to Rancho Santa Margarita). Just in these four districts, you and your predecessors have collected observations on 88% of the F/2020 pigeons and 71% of the previous pigeons. So, these should form four reasonably stable subpopulations for analyzing trends in pigeon demographics.
I broke the database down by time, as well. All told, you and your predecessors have observed nearly 7,500 pigeons and over 700 of their funny little courtships. Of these, about 7,000 were identified down to particular morphs. Of the identified birds, 4,100 were counted in the past and close to 1,100 by you. Looking at the distribution of individual flocks through time, I broke your predecessors' data into three time ranges: 2000 through 2003, 2007 through 2011, and 2013 through 2018. Team 7's job is to compare and contrast pigeon demographics (average size of flocks and average percentages of the seven morphs in them, first, in the two databases (coarse scale analysis) and, then, in the four Zip Prefixes by the four time periods (finer scale analysis).
The needed data have been preprocessed in the two spreadsheets:
- https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatchbySiteF20.ods (this semester's data)
- https://home.csulb.edu/~rodrigue/geog330/PigeonWatch/PigeonWatchDatabase00-18.ods (previous data)
Look for the Sort Zip tab at the bottom of the spreadsheets. When you open these tabs, you'll find data broken out by the four Zip prefixes, first, for the whole database in cells A365 to L383 of the 00-20 sites database.
Below that, you have everything broken out for the 2000-2003 time period from cell A385 through cell L403.
The 2007-2011 time frame is in cells A405 through cell L423.
The 2013-2018 time frame is in cells A425 through cell L443.
The F/2020 data are in the other spreadsheet for the 2020 sites, running from cell A100 through L115.
You'll need to gather and graph the following data:
- average flock size for the entire 00-18 database (row 374) and for the entire 2020 database (row 119). Turn that into a bar chart with just those two columns (past and present).
- Average flock size for Zip Prefixes 900, 907, 908, and 926 and the total for the four prefixes and the total for the entire database for each of the four timeframes: 2000-03, 2007-11, and 2013-18 from the 00-18 site database and for 2020 from the F/20 spreadsheet. Turn all that into a line graph in Calc, with time periods forming the X axis and average flock size the Y axis. Plot on that six lines for average flock sizes by time and space (900, 907, 908. 926, total for the four, and total for the entire database.
- Four stacked bar charts each with four columns of percentages, one column for each of the Zip Prefixes in each time period. Repeat for the next three time periods. If you choose a percentage stacked bar chart (far right option), you can easily see the differences in the dominant morphs from one region to the next. If you choose the middle option, you get the percentage information but the height of the overall columns reflects the size of the regional flock, so it's an easy way to track local change over time.
Do the sizes of the flocks by Zip Prefix change through time? Are the F/20 flocks noticeably smaller, perhaps due to COVID? Or is there a long-term trend in flock sizes that this semester is merely continuing? Are there shifts in the relative dominance of the various morphs across space or through time? If so, what might be going on (random fluctuation? change in the predatory environment? municipal pigeon eradication programs?)?