parallel = TRUE not used in optimStrata method = "spatial"

[nsaby]

Hello,
I'm using this code

solution <- optimStrata (
method = "spatial",
errors=cv,
framesamp=frame,
iter = 15,
pops = 10,
nStrata = 5,
fitting = c(0.9,0.9),
range = c(15000,15000),
kappa=1,
writeFiles = FALSE,
showPlot = TRUE,
parallel = TRUE,
cores = 10
)

But I get this output from the function:

Input data have been checked and are compliant with requirements
Sequential optimization as parallel = FALSE, defaulting number of cores = 1

I have version 1.5.4
Thanks for your help

[barcaroli]

The 'parallel=TRUE' parameter works only if more than one domain is defined, as parallelization applies to different domains at the same time. I suppose you used the function 'buildFrameSpatial' to prepare the frame, what values did you assign in the 'domainvalue' parameter?

[barcaroli]

If you want to speed up the processing with the spatial method, use the last version of SamplingStrata (1.5-5) available from GitHub: devtools::install_github("barcaroli/SamplingStrata")
In this last version, part of the optimization code has been translated into C++

[nsaby]

Thanks. Yes I defined only one domain in that exercise and used the function 'BuildFrameSpatial'

[nsaby]

I am wondering if it is possible to use the function "Bethel" with the weight of the strata equals to the relative area instead of the number of the populations strata. I am not sure I got it right. I need to run the bethel algorithm but with my own strata (not optimised). It is the 'N' column in the strata dataframe. Number of population units in the stratum
Maybe I should copy this question elsewhere?
Thanks in advance

[barcaroli]

If you consider the square kilometers as the units of your frame, you can assign the number of square kms to the N of each single stratum. The bethel algorithm will define the optimal allocation as the number of single kms to be selected in each stratum.

[nsaby]

Thank you for your time,
I'm in the process of designing a sampling plan to monitor the quality of French soils throughout the country. We are sampling soils using a site that corresponds to a 20 m x 20 m square. How is it possible to calculate the number of sites rather than the number of kilometres to be selected in each stratum?

[barcaroli]

Considering that France's total area is 551,695 km2 and your selection unit is 20x20 m2, if I am not wrong you have a population of 551,695,000,000/400=1,379,237,500 selection units. Indeed a lot. Anyway, the feasibility of the Bethel application depends on how you define the strata and how you calculate the mean and standard deviations of your target estimates in each stratum.

[nsaby]

I'm expecting to use the HT estimator based on a simple random sample within a stratum. The strata are defined using a map of soil and land use.

[barcaroli]

If you want to use the Bethel algorithm to optimally allocate your sample, you need to evaluate the variability of your target estimates in the strata you have defined. I understand that strata are characterized by the prevalence of soil characteristics and land use. You could fit a model between these available variables and your target variables, and estimate for each stratum mean and standard deviations of the target variables. If you do not have this possibility, then you can just use a simple sample size calculator (see https://www.calculator.net/sample-size-calculator.html) to determine the sample size in each stratum.

[nsaby]

Ah... yes sorry, I have answered too quicly... I am using maps of the target variables produced by a model (digital soil mapping). I used the eq of Ospats to estimate the standard deviation as in your package (option spatial).