Beginner to PC-ORD

[stephani...@eagles.usm.edu]

Hi all,

I am new to PC-ORD. I have been using excel up till now to manage my community data set but it is time for some real analysis.  I am looking to run a hierarchical cluster and NMS. I have managed to load my data set into the program from my .xls sheet but it doesnt seem to like how it's presented. I have my stations in column 1 and families in row 1.  It is density data instead of abundance.  Can anyone cue me in on any hints as to how I can run my analysis. I have printed a manual and I've been reading but I think I am missing something.  I have the instructions to run those particular analysis but they won't give me any results.  Help please!

Stephanie

[Bruce McCune]

Stephanie, it sounds like you need to import your data with File | Import | Simple Excel Spreadsheet.

Also, see the Getting Started | Data Preparation section of the built-in help system.

-Bruce McCune

--
You received this message because you are subscribed to the Google Groups "PC-ORD" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pc-ord+un...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
Content-Type: application/vnd.ms-excel; name=HC1.xls
Content-Disposition: attachment; filename=HC1.xls
X-Attachment-Id: d6835e62-ff21-4251-aff1-ba5d2c89e54f
Content-ID: <d6835e62-ff21-4251-aff1-ba5d2c89e54f>

[stephani...@eagles.usm.edu]

I am having technical issues with the program.  Im not sure if it is because I am using version 5 on windows 7.  I tried to use the help link but it will not open because of a windows 7 issue that is unsupported.  I have also reformatted my data in excel as suggested on the data preparation and saved as .xls however when I go to import the main matrix (using the excel selection) I get activation method of range class failed.  Im wondering if I need to go ahead and purchase the new PCORD 6 or if its just a problem with my data. 

Thanks for your help and suggestions.  

Below is a sample of how my data is arranged.

Stephanie Taylor

30	sites
37	families
	Q	Q	Q	Q	Q	Q	Q
	Antennar	Balistid	Bregmace	Caproida	Carangid	Chaetodo	Chloroph
site21	0	5.138451	0	0	183.9024	2.840326	0
site22	0	0	0	0	1.295756	0	0
site33	0	0	0	0	0	0	0
site34	0	0	0	0	9.085125	0	0
site35	0	0	0	0	1.75781	0	0
site36	0	0	0	0	13.65808	0	0
site37	0	0	0	0	35.44801	0	0
site38	0	0	0	0	120.2733	0	0
site39	1.372642	0	0	0	113.9293	0	0
site40	0	0	0	0	4.781753	0	0

[Susan Will-Wolf]

Stephanie,

I am using PCORD 5 on windows 7 with no problems, so I think your problem must be something else. I presume you have made sure you have all the v5 updates that are available.

Susan WW

> --
> You received this message because you are subscribed to the Google Groups "PC-ORD" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to pc-ord+un...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.

--
Susan Will-Wolf
Senior Scientist, Department of Botany
University of Wisconsin - Madison
430 Lincoln Drive, Madison, Wisconsin 53706-1381 USA
email: sww...@wisc.edu
phone: 001-608-262-2754 FAX: 001-608-262-7509

[stephani...@eagles.usm.edu]

Yes I believe I have the updates installed, but thank you for making me aware of them. I still have the problems described above.  I wish there was a way I could trouble shoot.  I was able to load my data(.xls) in the improper format but PCORD couldn't run anything of course.  However after I made my data as described above per instructions it wont load it at all.  I still get the Activate method of range class failed.

I appreciate any advice offered.  Thanks!

Stephanie Taylor

[Bruce McCune]

Stephanie, this is something simple in your way, but it's hard to say what from the info provided. I suggest contacting technical support at pcord.com and attaching your data set.
-Bruce

[stephani...@eagles.usm.edu]

I believe you are right.  I will definitely do that.  Thank you for the point in the right direction.  I'm really glad this group is here.  Ive found it very informative and supportive. 

Thank you

Stephanie Taylor

[stephani...@eagles.usm.edu]

Hello again,

I was finally able to solve the issue with importing my data on v5.  I am trying to run a NMDS with community data concerning ichthyoplankton.  I have both the community matrix and the environmental matrix.  The program will go through with the NMDS and I go to graph the data but what I'd really like to do is the joint plot which I chose but it will not impose the radiating lines from the second matrix.  It only shows the 2d graph. I have selected the features to get the graph to display the points properly etc under the option/preferences option.  Any thing I am missing to get radiating lines? I am trying to distinguish what environmental feature is most responsible for the particular layout of the NMDS or if there is a gradient. 

Thank you for any and all advice

Stephanie Taylor

[Peter Nelson]

Hi Stephanie,

It sounds like your second matrix (environmental data) variables don't
have strong enough correlations with your ordination axes to show up.
You can control this in a couple ways, once of which is to click the
little box at the top of the graphed ordination that defaults to .200.
That will allow you to decrease the correlation threshold. If none of
your second matrix variables have a linear correlation strong enough
to show up, you could try transforming them to make them more linear.
You could examine this in a couple ways, one of which is by clicking
the little box (still in the graphing mode) that has the little
scatter of points but says "2nd" at the bottom. You can then flip
through the bivariate graphs of each NMS axis and each second matrix
variable to look for potential candidates for transformation.

Hopefully I am understanding your question correctly and said
something useful.

Peter R. Nelson, PhD
Post-doctoral Researcher
Department of Botany and Plant Pathology
Cordley Hall 2082, Oregon State University
Corvallis, Oregon 97331-2902
Phone: 541-231-5584
http://peterrnelson.weebly.com/

[Bruce McCune]

Stephanie, if you are selecting the joint plot and nothing is showing up, chances are you have no vectors long enough to pass the criterion to plot. But you can change that -- click the button that says .200 on it and set it to a smaller value (such as .000). If you have Q variables in the second matrix, you should see some vectors.
-Bruce McCune

[stephani...@eagles.usm.edu]

Thank you to you both.  I did indeed have to change the default because my environmental variables were not strong enough, but I even chose those variables with the strongest relationship in the first place.  I am now moving onto Beal's smoothing and then run NMDS.  I am also checking my Diversity, richness, and evenness calculations that I have done through excel.  I have used row and column summary but where can I find beta diversity?

Thank you for your help.
Sincerely,

--
Stephanie M. Taylor
Graduate Assistant
Gulf Coast Research Laboratory
The University of Southern Mississippi
703 East Beach Drive
Ocean Springs, MS 39564
228-238-7944

[Bruce McCune]

Stephanie, for beta diversity statistics run Advisor | Show Current Profile.
Bruce

[stephani...@eagles.usm.edu]

Ah ha! Much appreciated.  I feel like I am finally beginning to understand the program.  I have more analyses to run.  Again, Thank you for your help.  I'm sure I will be back. 

[stephani...@eagles.usm.edu]

I have a question on beta diversity.  I have been using density data to do my general analysis.  I proceeded with NMS on this data, and it was very crowded so I decided to log transform it and then run the NMS.  I am trying to also acquire the beta diversity values, but now I am having a problem of choosing which format the samples should be in order to get a value that is reliable.  Should I use the raw abundance data, density data, or the log transformed values? Also should I delete rare species before I do this as well? I have already gotten the diversity, richness, and evenness without deleting rare species but I was unsure of the procedure for beta. 

Thank you for your time

Sincerely

[stephani...@eagles.usm.edu]

Also, one more question for this evening.  Where would I find the Monte Carlo test result for the probability of finding similar final stress obtained by chance if I am running NMS? I am trying to write up my findings.  I realize that the p under the Monte Carlo portion of the NMS results relates to the proportion of variance for each axis. 

Thanks!

[Bruce McCune]

Stephanie,  You can use beta diversity in different ways, and how you are using it affects what you should report. Ultimately it is a property of a sample, rather than an innate biological quantity. As a descriptor of the whole community, I suggest calculating it before removing rare species. As a descriptor of a data set, including the inherent difficulty of representing it in an ordination framework, I recommend calculating it for whatever matrix you have chosen for your analysis (might be transformed and with rare species deleted).
-Bruce

[Bruce McCune]

Stephanie, the p values in the NMS output refer to the final stress, not any proportion of variance.
-Bruce

[stephani...@eagles.usm.edu]

I am in need of some further assistance.  I am finally running the NMS with the log transformed values.  I have reached several with no ordination was found.  I have already looked for outliers and eliminated some in one case, and none in another.  I have run these same values in a PCA, but I decided that it was not appropriate for community data since the skew and kurtosis were >1.  The PCA produced nice results.  I have also run NMS with this data as density only (no log transformed) but the data was too congested to get a clear image.  I thought if I transformed the values it would help to spread the data out.  Please advise. 

[Bruce McCune]

Stephanie, a couple of responses on this:

- first, why was no useful ordination found? There are several potential reasons: stress too high? failed randomization test? (see potential causes at bottom of p. 133 in McCune and Grace (2002). Some ordinations that fail the randomization test might still be usable. You can force NMS to keep the best solution by turning off autopilot and the randomization test. See tips in the built-in help system on running NMS with autopilot off.

- A log transformation of won't necessarily increase or decrease the uniformity of spread of points in an ordination. But if there are very large and very small values in the same input matrix (say densities ranging from 1 to 10000, it is likely to improve the ordination. I suggest reading (or rereading) chapter 9 in McCune and Grace. Also, try following the "general procedure for data adjustments" on p. 78.

Good luck,
Bruce McCune

[stephani...@eagles.usm.edu]

Bruce,

Thank you for your response.  I found it very helpful.  I had to turn off the autopilot since my stress was ok but not great.  I decided to go with the data set that had lower stress and solution was found.  I saved the result file but I accidentally deleted the other two point files that go with the NMS result.  Is there anyway to regenerate that information for that particular NMS result so I can bring those points into excel?

Stephanie Taylor

Again, Thank you for your patience and time :-)

[Bruce McCune]

Stephanie, the graphrow and graphcol files contain the coordinates for the points for graphing. If these are gone but you still need them you'll need to rerun NMS to regenerate them.
Bruce

[stephani...@eagles.usm.edu]

I was fortunately able to find these coordinates at the end of my results file that I saved as well as recover the graphrow and graphycol. 

I am trying to rerun a particular NMS. I have the seed for the random number generator and I have turned off randomizations. I am able to reproduce the result but not the same graph result.  I am running it in autopilot but I think it is using the seed since I am getting the same stress results, but not the same iterations.  Am I missing something?

 As always thank you for your help

Stephanie Taylor

Graduate Assistant
Gulf Coast Research Laboratory

[Bruce McCune]

Stephanie,

This is very common that different runs will end up with similar (or even exactly the same) stress, but differ in some details of the graph, or differ in some trivial geometry of the graph (e.g. rotation or reflection of one result into another).

If you are running autopilot, it generates new random number seeds each time. But it is still common to get the same stress resulting.

Try rotating (Rotate | by angle continuous) or flipping (Scaling | Reflect) your graph, and you should be able to come close to your previous configuration.

Bruce McCune

[stephani...@eagles.usm.edu]

Bruce,

Thank you. I will give it a try. I can see that the results are indeed similar but its usually the vectors that are most different.  If this is trivial then I probably do not need to be worried that I cannot reproduce the exact same graph?

[stephani...@eagles.usm.edu]

I have 2 more questions.  In the NMS results file where would I find the proportion of variance explained by each axis?  I have been looking online and in Analysis of Ecological Communities but I am missing on where to find this information so I can report it in the NMS results. 

Second, Does the output for the NMS for graphing include the vectors for the second matrix? I found the graph.col and graph.row and the points it contains for graphing the NMS result, but where would the vectors for the biplot be?

Again thank you for your patience and help

[Bruce McCune]

Stephanie,

After any ordination, including NMS, you can evaluate the relationship between the original distance matrix and the distances in the ordination diagram by clicking on the R2 button in the graph view. Note that if you are rotating or reflecting your axes, you should do this calculation after the rotation.

The coordinates for the joint plot vectors are calculated within the graphing module, since they are after-the-fact, not part of NMS per se. In the graphing module you can get those coordinates by File | Save scores as | Joint plot tips.

Bruce McCune

[stephani...@eagles.usm.edu]

Bruce,

I was successful in making a similar NMS plot.  Thank you for your patience and instruction.

 I am running another NMS with different data set.  I have read in Analysis of Ecological Communities several times on NMS and autopilot but I guess I am misunderstanding exactly what the difference is between running an NMS w and w out autopilot besides the random seed generator?  I am very confused as to why the plots and stress are so different if I am rerunning the same data set with the same settings as the autopilot. 

Thank you in advance

Sincerely,

[Bruce McCune]

Stephanie,

If you run NMS with autopilot off, then you can choose all the options for the analysis. But with autopilot on, PC-ORD chooses everything for you, except for the distance measure and the tie handling option. User-provided random number seeds and all other user-set options are ignored with autopilot.

When you say "same settings as the autopilot" -- there are many different settings that you would have to check to be sure they are the same. If even one of these is different, then it could make a difference in the results. But with most data sets the actual difference will be small and the same basic structure will emerge over and over.

If you want to check how big the difference actually is, without regard to rotation or mirror image configurations, try this. For each of two ordinations, save the graphrow.gph files under new names. Then run Ordination | Compare Scores, and it will give you a statistic for how similar the interpoint Euclidean distances are between one ordination and another.

-Bruce