Recent Changes

Re: Other sequential organization for lectures? — by Helene Wagner — last modified 2009-03-11 12:01

We could probably come up with many possible orders and not agree on any. However, now that we see the draft slides, we should give this some thought again. Thanks Yessica for the proposition! I'd like to throw in the following suggestions: 1. Lecture 13 would fit well after lecture 3 or 4 and before lecture 6. 2. Lecture 9 would profit from being after lecture 11. 3. Hence a possible sequence would be 2 3 4 13 6 7 5 8 12 11 9 14 10 Sorry this looks pretty jumbled! Helene

Re: About graph and link between lecture 9 and 4 — by Helene Wagner — last modified 2009-03-11 11:56

I also have a slide on graph theory / network analysis in lecture 5.  I can turn this into a link to the previous lecture.

Link to model selection — by Helene Wagner — last modified 2009-03-11 11:53

This lecture relies on model selection, so maybe it should come after lecture 11?

Black bear example — by Helene Wagner — last modified 2009-03-11 11:43

Hi Sam - The black bear example is used in several lectures (8, 11, 12), we should think about exactly which slides go where. Here you already show the causal modeling, but this is dealt with in a later lecture. Is it meant as an outlook?

Spatial autocorrelation — by Helene Wagner — last modified 2009-03-11 11:39

I wonder whether spatial autocorrelation could / should be moved to the distance-based methods lecture. Also, I have some reservations about interpreting significant autocorrelations at large distances: this requires a good explanation (the example shown is ok in that respect) but also that this is not an effect of very few observations. For any distance class beyond half the maximum distance of the sample, not all observations will contribute. As you get to the largest distances, very few observations are included in the calculations. You may have 10 pairwise comparisons, but they all are between the same two patches... While Moran's I is focused on testing for autocorrelation, variogram based methods are more directly interpretable (the y axis is a measure of genetic distance) and easily related to pairwise fst plots etc. while allowing for testing for autocorrelation. Maybe these could be mentioned too (e.g. Wagner et al. Genetics)?

Sampling design — by Helene Wagner — last modified 2009-03-11 11:32

Where do we treat sampling design, sample size, and the problem of unsampled populations within the study area? I thought this would be treated in lecture 3.

random vs. IBD — by Helene Wagner — last modified 2009-03-11 10:07

Looks good! Although this is already implied, it may be worth, to include a slide highlighting that naively we would use randomness as a null model, as is done frequently in most areas of statistics and in population genetic models (panmictic populations). One would then test for IBD as significant departure from randomness. In Landscape genetics, however, IBD is the starting point and needs to be built into the null model so as to test for significant departures related to landscape.

Re: Assignment tests: lecture 3 or 4? — by Melanie Murphy — last modified 2009-03-11 08:42

I think assignment tests can be covered well in lecture 4.  One thing that would be very helpful in lecture 3 would be a slide/table with assumptions of the different distance measures (HW assumptions, other, no assumtions made) and what that are based on (heterozygosity, alleles, or allele frequencies).  It is crucially important that students start thinking about what these distance measure mean and the underlying assumptions so they can make informed decisions with their own data.  I would also like to see proportion of shared alleles added.  Both Sam and I talk about this distance measure later in the seminar, so it would be helpful if it was introduced/compared in this lecture. 

Re: Literature lists for each lecture — by Melanie Murphy — last modified 2009-03-11 08:23

I think that sounds like a good plan.  Having a list of references cited in the talk is important.  I also think that having a broader list of references that more broadly covers the topic would be invaluable to the students (partially take the place of a book).  Maybe accompanying each lecture there could be a file that would list something like this: Required Readings: XXXX Additional Recommended Readings: XXXXXXX Literature Cited: XXXXX Additional Relevant Literature: XXXXX

Assignment tests: lecture 3 or 4? — by Helene Wagner — last modified 2009-03-11 04:26

Lecture 3 seems really full, and the topic of assignment tests only makes sense in the context of discrete populations, hence it may be a good idea to move that topic to lecture 4?

Re: Literature lists for each lecture — by Helene Wagner — last modified 2009-03-11 04:09

Here are my two cents: I think we should distinguish different types of literature listed: - required reading for a lecture (ca. 2 per lecture, reviews that provide introduction for non-specialist, or key examples used e.g. in student activities), may include additional commented list of further recommended readings. - literature cited in a lecture - anything else about a topic as suggested by Melanie. This does not need to be tied directly to a specific lecture.

Re: Extra topic suggestion: Difference among genetic markers — by Helene Wagner — last modified 2009-03-11 04:01

I think the topic is introduced in lecture 2 and the recommended reading for that week. What exactly do you think is needed in addition? Would the choice of markers with the whole scale question be a suitable topic for a group project? We are looking for group projects that write a review-type paper on an open question of methodological choice at a specific step in the process of a landscape genetic study.

Re: Extra topic suggestion: Difference among genetic markers — by Melanie Murphy — last modified 2009-03-09 06:56

I agree that it would be great to briefly talk about this in one of the early lectures.  In addition, it would be nice to talk about the scale of these genetic markers (temporal (evolutionary) scale and spatial scale) and the need for research investigating the interaction between marker scale and ecological scale. 

Other sequential organization for lectures? — by Yessica Rico — last modified 2009-03-09 06:43

I don't know if it is possible to change the sequence of the lectures, but reviewing the materials I have a suggestion for that: the lectures for week 5, 8 and 12 are related with topics very linked since the first is about different views of landscapes, the second introduces the topic of distance measures related with landscape features and the last one cover more deeply the importance of including the landscape perspective in the response organism. In this order (5,8,12), the example of the mountain goat maybe could be best covered only during the slides of week 12.     <br /><br />In addition, following these three lectures, week 11 could be the next since is related with model selection and validation that put as in example the bears study introduced previously in week 8 and 12. Also, the lecture for week 7 could be covered after all of these previous lectures because it talks about simulation modeling related with IBD models, so the students at this point of the course are more familiar with distance measures related with landscapes.<br /><br />Overall, my lecture order suggestion is: week 2, 3, 4, 5, 8, 12, 11, 6, 7, 9, 13, 14, 10. <br /><br /><br />

Extra topic suggestion: Difference among genetic markers — by Yessica Rico — last modified 2009-03-09 05:59

I have a suggestion for an extra topic about the correct identification of genetic markers in population genetics studies (when to choose dominant markers vs co-dominant, difference between organelle and nuclear markers, etc). Because it is probably that the course will be taken by people with different backgrounds and expertise it would be very useful and more integrative for the course to talk about why is fundamental to choose the proper genetic marker for which research question. This topic could be included briefly in the lecture slides of the second week (genetic diversity) or third week (gene flow).

Citation of figures, graphics, and pictures — by Melanie Murphy — last modified 2009-03-08 17:48

Many of the slides in the lectures are fantastic with slick figures, graphics, and photos.  Because these lectures will be distributed to multiple universities and internationally, we should all go through our slides and make sure that 1) all figures/results are clearly sited to the appropriate paper (or as modified from a source), 2) any graphics not generated by the author are cited (web page, other reference), and 3) credit for any pictures not taken by the lecture author(s). 

Re: False discovery rate — by Helene Wagner — last modified 2009-03-08 17:05

Hi Stephanie - yes, could you please send me the paper? Thanks, Helene

Allele Matrices - need correction — by Melanie Murphy — last modified 2009-03-06 16:17

Hi Sam - I like how you are demonstrating genetic distance by having students take a look at the matrices.  However, the matrix you use as an example is inaccurate for representing microsatellite data.  There will be two alleles/locus for a diploid species.  No matter how alleles are distributed amoung loci, this is not possible with the current example.  In addition, due to  this dependency A = B in the case of the Bray-Curtis (# of loci * 2, or more generally # loci * ploidy).  I put an example below, keeping the number of alleles (6) the same as your example.  Each letter represents a locus (X locus, 3 alleles; Y locus, 2 alleles; Z locus, 1 allele).  I suggest updating your matrices and indicating loci in some way (color, heavy outlines, Loc1_A/Loc1_B, etc).  AFLP data could also be represented, but would be presence/absence of alleles (0,1).                X.1    X.2     X.3    Y.4    Y.5    Z.6       A    B    C Bear1         1       1      0        2       0      2        6      Bear 2        0       1      1        1       1      2              6    Min             0       1      0        1       0      2                     4

Re: Modifications — by Rolf Holderegger — last modified 2009-03-06 07:26

Dear Stéphanie. I will include your comments into the lecture this weekend. Thank you very much, Rolf

Re: Overlapp — by Rolf Holderegger — last modified 2009-03-06 07:24

Dear Stéphanie, I actually did this on purpose, so that students would have been reminded of something they had already seen. Do you think this is problematic? Rolf

Some comments on distances — by Rodney Dyer — last modified 2009-03-04 08:51

Sam, Excellent slides. Without hearing your presentation with them (my caveat), there are two items that I thought may help out. -  While it is true that people use Fst as a measure of differentiation, it is not metric (e.g., there are cases where it fails the triangle inequality) and as pointed out by Wright (1978) that "The fixation index is thus not a measure of the degree of differentiation in the sense implied in the extreme case by the abasence of any common allele.  It measures differentiation within the total array in the sense of the extent to which the process of fixation has gone toward completion." - While it is true that the Mantel is used due to non-independence, the verbage on the slide suggests that none of the items in the matrices used for the Mante are independent when in actuality, you can get independent samples from a pair-wise matrix (just not too densly). Rodney

Re: About graph and link between lecture 9 and 4 — by Melanie Murphy — last modified 2009-03-03 22:54

Dear Stephanie - I think the population assignment lecture should definitely come before the network lecture.   You are right, several of the population assignment methods use some underlying form that is considered a graph.  However, these approaches are really using them for interpolation of genetic variation (identify major breaks, etc) and not really as a network.  In addition, they fit nicely with population assignment information.  To link the lectures, you could introduce these examples as interpolation methods that can also be used as networks.  I could then reinforce those that were given as interpolation methods in your lecture.  It could be that this confuses the issue.  If you like, we can converse in more detail via e-mail to figure out a good way to connect the lectures. There is overlap/synergy between graph theory and network analysis.  The best explanation of the difference in my opinion is that networks are some sort of graph (named or unnamed), but not all graphs are used for network application.  A network has some flow (implicitly or explicitly modeled depending on the approach), as in electricity, goods, movement etc.  Graphs can be used for many applications, including theoretical mathematics.   In addition, graph theory can be used to build a graph/network.  However, in ecological application there is a tendency to use graph/network interchangeably. I will go back through my slides and make sure I am consistent.  In addition, in two of the examples (circuitscape and gravity models) flow is explicitly in the equation.  I will also go through the wording on the types of models and see if I can do any clarification.  Thank you for the feedback.  Let me know if we should talk further on how to link the lectures. Melanie Previously Stephanie Manel wrote: Dear Melanie, I'am wondering if the lecture 4 has not to be after lecture 9, since a lot of the method used to idendify discrete populations are based on populations graph : Monmonnier algorithm and TESS.  If we do not change the order, we need  to have a link between the two lectures. Question: I think that I miss something about network method and perhaps my question is very naive but I'am not sure to be clear with network methods: are all the method based on a graph theory a network method? Or what make the difference between a network method and a method based on graph theory? Finally are Monmonier algorithm and or TESS network methods? Stéphanie

Proportion shared alleles - equation and matrix need correction. — by Melanie Murphy — last modified 2009-03-03 21:50

The D in the equation for proportion of shared alleles is the number of loci and not number of alleles.  This is especially important for markers (such as microsatellites) where loci are the unit of observation, and not alleles.  Also the matrix should be filled with frequency of the allele in the individual (0, .5, or 1) and not count (0, 1, 2). 

False discovery rate — by Stephanie Manel — last modified 2009-03-03 04:30

I suggest to introduce false discovery rate in the lecture that has been defined as more robust than p-value to detect fasle positive and to be use in multiple testing. I can send papers about FDR if necessary. Stéphanie

About graph and link between lecture 9 and 4 — by Stephanie Manel — last modified 2009-03-03 04:01

Dear Melanie, I'am wondering if the lecture 4 has not to be after lecture 9, since a lot of the method used to idendify discrete populations are based on populations graph : Monmonnier algorithm and TESS.  If we do not change the order, we need  to have a link between the two lectures. Question: I think that I miss something about network method and perhaps my question is very naive but I'am not sure to be clear with network methods: are all the method based on a graph theory a network method? Or what make the difference between a network method and a method based on graph theory? Finally are Monmonier algorithm and or TESS network methods? Stéphanie

Overlap? — by Stephanie Manel — last modified 2009-03-03 03:45

-I also mention the paper of Latch et al in the lecture "identifying discrete populations". We need to check together that there is no overlapp and a link between the both lectures. Stéphanie

Overlapp — by Stephanie Manel — last modified 2009-03-03 02:33

At the moment I find a very small overlap between  the lecture of this section and the lecture 14 about adaptative variation: the same figure from Joost paper is used in the two lectures. However, not sure that it is a problem...It depends how it is intepreted.

Modifications — by Stephanie Manel — last modified 2009-03-03 01:57

Rolf, I send you by e-mail an annoted version of your document and the recent paper of  of Schmidt et al. (2008). I was not able to put it in "your" directory. Stéphanie

Literature lists for each lecture — by Melanie Murphy — last modified 2009-03-02 11:13

I think it would be a huge benefit if each lecture was accompanied by a list of literature on the topic.  This literature list would include those sources cited in the lecture, as well as a fairly complete list of other publications not covered during the lecture.  The combined references would be a huge resource to students and others.  These lists can also be more organic, and we can add to them over the next year as new research comes out before the course is presented.    Any thoughts? 

Re: Formats — by Lisette Waits — last modified 2009-02-26 20:44

Okay...I'd upload doc or pdf only.  Were the powerpoint formats okay? Previously Rodney Dyer wrote: Would it be possible to have the documents in a format other than *.docx?  I imagine some of the students and other participants are not Windows users and cannot read these formats.  Regular old *.doc or *.pdf will possibly have the largest audience. Rodney

Re: Formats — by Melanie Murphy — last modified 2009-02-25 11:41

Maybe we can have a standard of using pdf for all files that are going out to students (slides, documents, notes, etc).  <br />

Formats — by Rodney Dyer — last modified 2009-02-25 08:48

I posted this elsewhere but am reposting it here for general consumption.  Can we try to use open formats for the materials being presented?  There are many of us as well as the students that we are going to be reaching that will not be able to open *.docx or *.pptx files.  I imagine *.doc, *.pdf, and *.ppt would be most efficient in reaching all of our audience. Rodney

Formats — by Rodney Dyer — last modified 2009-02-21 17:12

Would it be possible to have the documents in a format other than *.docx?  I imagine some of the students and other participants are not Windows users and cannot read these formats.  Regular old *.doc or *.pdf will possibly have the largest audience. Rodney