Problems with spatial autocorrelation plot

[Gabriella Scatà]

Hi,

I have just started using the gl.spatial.autoCorr() function and I already have issues at the start.... :/

The functions runs fine, but the plot looks weird, no matter which distances (max and bins) I use.

I don't get any curve/lines representing the relatedness of samples across bins, I only get a bar for each population or for each sex (I tried both with pop = pop, or pop=sex) that is fixed at the start...

What am I doing wrong?

I checked my lon and lat info:

it is reported correctly i think:

head(my_dataset_genlight@other$latlon)

ind ID   lat          lon

1          -25.5     153.35

I tried also to convert the lat lon data with Mercator:

the Mercator function was giving me this warning message:

 # Warning message:
    # In .pointsToMatrix(p, checkLonLat = FALSE) :
      # Suspect column names (longitude and latitude reversed?)

So, I reversed the order of lat lon in my genlight, like this

my_dataset_genlight@other$latlon = 

   my_dataset_genlight@other$latlon[ ,c(2,1)]

then run Mercator:

my_dataset_genlight@other$latlon = Mercator(

my_dataset_genlight@other$latlon , inverse= T)

ind ID     lat              lon

1          -0.00022     0.0013

Although I am not sure if I should transform to Mercator, if my lat and lon is taken from Google map positions?

Anyways, this is the code I used:

gl.spatial.autoCorr(my_dataset_genlight,
                    bins = seq(0, 1500000, 100000),
                    plot.pops.together = FALSE,
                    plot.colors.pop = c('#E69F00', '#5D3FD3', "red", "blue", "green", "violet"))

I also tried running the functions in this way:

- plot.pops.together = TRUE

- replace pop(my_dataset_genlight) = my_dataset_genlight@other$ind.metrics$sex

- changing the bins to seq(0, 150, 10)

But I always get the same result with no lines plotted...and a single bar for each pop/each sex at the very start of the plot...

I also tried to convert to Mercator, but with inverse = F (which seems more correct as I am passing from latlon to Mercator), but then when I run the spatial autocorrelation with the dataset converted in this way, i get the message:

Error in x_temp@other$latlon[ , c("lon", "lat")]: suscript out of bound

What am I doing wrong?

See plot attached...

I would really appreciate some help!!

Thanks a lot!

Gabriella

[Jose Luis Mijangos]

Hi Gabriella,

The problem is that the spatial autocorrelation analysis is performed by population, as mentioned in the function documentation:

"If the input is a genlight object, the function calculates the linear distance for Dgeo and the relevant Dgen matrix (see Dgen_method) for each population"

In your case, all the individuals within a population have the same GPS location. 

One solution would be assigning the same population to all the individuals, as shown below:

library(dartRverse)
t1 <- your_genlight
pop(t1) <- rep("pop",nInd(t1))
res <- gl.spatial.autoCorr(t1)

Cheers,

Luis 

[Gabriella Scatà]

Oh, I see, yes that makes sense!

Buy then can I also ask this again for clarifications:

- do I need to convert my Google maps coordinates lat lon coordinates (i.e. -18.00, 152.30)  into Mercator or can I just use my Google maps coordinates? In the function documentation, it seems I can use both, is that correct?

- but do I need to convert to Mercator if I want my bin sizes and distance information in the plot in meters? Or does the function automatically take that into account the different type of information and does the conversion to meters before plotting?

- In case I convert to Mercator - it seems I need to reverse the order of genlight@other$latlon to "lon lat" to use Mercator() (correct?) --> in that case, do I need to reverse the order back again to "lat lon" to use the genlight in gl.spatial.autoCorr()? 

- To decide the bin sizes and min -max distance to plot, can I just look at my max range of distances? so say if my sampling goes from a certain location to 1500 km away, is it correct to code it in meters in the gl.spatial.autoCorr() function? --
--> like this: bins = seq(0, 1500000, 100000) ?

Thanks a lot for your feedback once again!

Best,

Gabriella

--
You received this message because you are subscribed to the Google Groups "dartR" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dartr+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dartr/c3a5d6b5-46d6-4ab2-8d82-c2035ea79175n%40googlegroups.com.

[Carlo Pacioni]

Hi Gabriella,

you don't need to transform your coordinates, or rearrange the order of your lat and long. The function will correctly use Mercator and calculate the distances (which will be in metres) behind the scene. You can use any range intervals you want. That really depends on the questions you are trying to answer (for example, if you are exploring at what distance connectivity drops and you have no previous knowledge, what you suggest makes sense. Alternatively, if you want to test whether you have movements within sites, between sites and between regions, then you may want to set distance classes that reflect this structure) and whether, with the distance classes you use, you have enough data for these to be meaningful. Note that the second column of the table that is returned reports the number of pairwise comparisons. If you have only a few of those, your estimate is probably not accurate. I hope this helps!

carlo