This update was focused on enhancing SSN2 alignment with
spmodel.
"wave" tail-up and tail-down covariance types.
The wave covariance has oscillatory behavior that decays in magnitude
with distance.conditional() for ssn_lm() and
ssn_glm() model objects, which draws simulated values from
the distribution of newdata conditional on observed
data.ranger, randomForest, or
xgboost) account for spatial covariance.
ssn_decorrelate() estimates decorrelation parameters via a
grid search evaluated on training/test splits or cross-validation and
fits a machine learning model to the decorrelated data;
ssn_decorrelate_data(),
ssn_decorrelate_grid(),
ssn_decorrelate_newdata(), and
ssn_recorrelate_newdata() support the underlying
transformation and grid search steps; and predict(),
print(), and tidy() methods are provided for
ssn_decorrelate() model objects.ssn_lm() model objects, for t-based (rather
than asymptotic z-based) inference in small samples.
satterthwaite(), which computes Satterthwaite
ddfs for the fixed effect coefficients of an ssn_lm() model
object.ddf argument to ssn_lm() and
anova() that controls whether Satterthwaite or asymptotic
degrees of freedom are used for hypothesis tests. In
ssn_lm(), if n (the sample size) is less than
or equal to 500, Satterthwaite ddf are computed and used, which implies
a breaking change in summary() and tidy()
output.confint() for confidence
intervals.emmeans::joint_tests() with
Satterthwaite ddf."cov", "ssn", "tailup",
"taildown", "euclid", "nugget",
and "randcov" values to the type argument in
vcov() for ssn_lm() model objects when using
Satterthwaite ddf."weight" value to the type
argument in predict(), which returns the prediction (i.e.,
Kriging) weights.kcv() for
ssn_lm() and ssn_glm() model objects.
kcv() generalizes loocv() (leave-one-out cross
validation) by holding out k (approximately equally-sized)
folds instead of single observations.reltol) passed to
stats::optim(method = "Nelder-Mead", ...) from
1e-4 to 1e-6, affecting ssn_lm()
and ssn_glm() model objects. The intent of this change is
to help prevent convergence to a local maximum that is not a global
maximum. This change may affect default backwards compatibility of
fitted models, depending on the shape of their objective function. If
the fitted model has changed due to the change in reltol,
adding control = list(reltol = 1e-4) as an argument to
ssn_lm() and ssn_glm() will reproduce the
original fitted model.eacf argument to Torgegram() to
compute the empirical autocovariance function.range_constrain argument to
ssn_lm() and ssn_glm() to constrain the range
parameter to enhance numerical stability. The default for
range_constrain is FALSE, implying the range
is not constrained.ssn_lmRF() functions to fit random forest
spatial residual models.
ssn_lmRF().predict() to
perform prediction.ssn_lm() and
ssn_glm().n / 2 * log(2 * pi) so likelihood-based statistics are
comparable on the standard scale. This is a breaking change that affects
backwards compatibility.spmodel.spmodel.spmodel.amongSitesBigDistMat() function from the
list of exported functions (i.e., it is now only able to be called
internally).spmodel.SSN2 website
titled “Adding Explanatory Variables to ‘SSN’ Objects Directly in R
Using StreamCat”.SSN2 website
titled “Fitting ‘SSN’ Models to Large Data Sets and Making Predictions
(i.e., Kriging)”size argument to the
local argument in ssn_lm() and
ssn_glm() from 100 to 200. This enhances the
approximation’s accuracy but can slightly increase computational
complexity.size argument to the
local argument in predict() and
augment() from 2000 to 4000. This enhances the
approximation’s accuracy but can slightly increase computational
complexity.local was specified (for large data sets) in
ssn_lm() or ssn_glm().local_index in
ssn_lm() or ssn_glm() model objects.spmodel version at least
0.7.0 so that AUROC() methods are properly
called.partition_factor from working properly when
non-NULL (#32).emmeans R
package for estimating marginal means of ssn_lm() and
ssn_glm() models.ssn_create_bigdist() function was added to create large
distance matrices using the filematrix R
package. Estimation for large data sets is performed by leveraging the
local argument to ssn_lm() and
ssn_glm(). Prediction for large data sets is performed by
leveraging the local argument to predict()
(and augment()). When local is used,
SSN2 looks for distance matrices created using
ssn_create_bigdist().ssn_import() so that it does not force an
overwrite of the netgeom column when it already
exists.verbose argument to ssn_import(),
ssn_import_predpts(), and createBinaryID() to
control whether warning messages are printed to the R
console.na.action argument to
predict.ssn_lm() and predict.ssn_glm()
functions to clarify that missing values in newdata return
an error.type argument in augment() for
ssn_glm() models to type.predict to match
broom::augment.glm().augment() for ssn_glm() models now returns
fitted values on the link scale by default to match
broom::augment.glm().type.residuals argument for
ssn_glm() models to match
broom::augment.glm().logLik() to match lm() and
glm() behavior. logLik() now returns a vector
with class logLik and attributes nobs and
df.AIC() and BIC()
from stats and removed SSN2-specific
AIC() methods.warning argument to glances() that
determines whether relevant warnings should be displayed or not.glances() about interpreting
likelihood-based statistics (e.g., AIC, AICc, BIC) when a one model has
estmethod = "ml" and another model has
estmethod = "reml".glances() about interpreting
likelihood-based statistics (e.g., AIC, AICc, BIC) when two models with
estmethod = "reml" have distinct formula
arguments.glances() about interpreting
likelihood-based statistics (e.g., AIC, AICc, BIC) when two models have
different sample sizes.glances() about interpreting
likelihood-based statistics (e.g., AIC, AICc, BIC) when two models have
different family supports (which can happen with ssn_glm()
models).cloud argument to Torgegram() to
return a cloud Torgegram.cex to
plot.Torgegram().Torgegram(); see
the robustargument to Torgegram()AUROC() function to compute the area under the
receiver operating characteristic (AUROC) curve for ssn_glm
models when family is "binomial" and the
response is binary (i.e., represents a single success or failure).type argument to loocv() when
cv_predict = TRUE and using ssn_glm() models
so that predictions may be obtained on the link or response scale."terms" prediction for
ssn_lm() and ssn_glm() models.scale and df arguments to
predict() for ssn_lm() models.dispersion argument to predict() for
ssn_glm() models.anova(model1, model2)) when
estmethod is "ml" for both models #25.anova(object1, object2) when the name of
object1 had special characters (e.g., $).ssn_glm() models when family = "beta" (#23).reexport.Rd to reflect changes in
spmodel v0.8.0’s handling of AIC() and
AICc().README.md) updates as part of a submission to Journal
of Open Source Software. Relevant issues associated with the review
are available at #11, #12, #13, #14, #15, #16, #17, #20, #21. The review is linked
here..ssn
folder that is accessed when importing SSN objects via
ssn_import().ssn_names() to return column names in the
edges, obs, and preds elements of
an SSN object.Matrix::rankMatrix(X, method = "tolNorm2") to
Matrix::rankMatrix(X, method = "qr") to enhance stability
when determining linear independence in X, the design
matrix of explanatory variables.X
has perfect collinearities (i.e., is not full rank).format_additive argument from
ssn_import() because of transition to geopackage support,
which eliminates the need to convert additive function values to
text.create_netgeom() function to create the
network geometry column for the edges, obs,
and preds elements in an SSN object.SSN_to_SSN2() that caused an error using
ssn_write() with no prediction sites.names.SSN() with ssn_names(), as
names.SSN() prevented proper naming of elements in the SSN
object.netgeometry to
netgeom to avoid exceeding the 10 character limit for
column/field names while writing to shapefiles (#2).family is missing in
ssn_glm() (#8).SSN_to_SSN2().Torgegram() that prevented intended
computation when cutoff was specified.plot.Torgegram() that occasionally
prevented proper spacing of the legend.ssn_glm() model objects (and their
summaries) when all covariance parameters were known.euclid_type
was "none".taildown_type was "spherical".ssn_glm() objects.