NEWS

---

spatialFDA 1.3.3

• rewrote summary.functionalGam and added a new coef.functionalGam to have no breaking changes in spatialFDA due to changes in refund::pffr naming.
• this makes spatialFDA again compatible with the OSTA chapter.

spatialFDA 1.3.2

• covariance matrices are now corrected for cluster-correlations along the domain $r$ with sandwich-estimators by default - thank you @fabianscheipl for implementing this in refund::pffr
• models can now also be fit with gamm4
• Experimental: implemented an AR(1) penalty along the domain $r$ in bam fits. This is to some extent already adressed by the cluster-robust sandwich estimators so might be removed in the future.

spatialFDA 1.3.1

• option to square-root transform the weights in the functional Gam
• changes to cutoff calculations at the saturating end of $G$ functions
• fix as.formula calls by passing an empty.env call
• cleaned up spatialInference from exploratory code.

spatialFDA 1.1.22

• fix sample_id call if model is a GAM instead of GAMM.
• added the option to define a ridge penalty parameter which is added to the matrix $H$ as defined in mgcv::gam. In order to do this we need the number of parameters p from a prefit of the model.

spatialFDA 1.1.21

• allowing for the explicit passing of correction in spatialInference and crossSpatialInference to fix issues with calculating Lcross.
• passing messages with verbose logical for better usage
• bug fix in tests with removed subsetby argument

spatialFDA 1.1.20

• only returning the transformed curves in spatialInference plus removing the assays and the rowData for faster computations.

spatialFDA 1.1.19

• fixing plotFbPlot ylims across conditions to have the same range-

spatialFDA 1.1.18

• removed the argument subsetby from spatialInference and crossSpatialInference as this will always be the same as image_id in this function
• indicated in the documentation that ... in spatialInference and crossSpatialInference is passed to both spatstat.explore functions and to refund::pffr.

spatialFDA 1.1.17

• added an assertion that the maximum radius considered must always be smaller than the maximum window length of each image.
• added a test for this case
• add a heuristic to check for a convenient and good rMax to consider
• added heuristic to the vignettes

spatialFDA 1.1.16

• adjusted that all failed calcMetricPerFov output NA in the dimension of rSeq, clearly stating which values in metricRes where NA.
• return the raw non-filtered metricResRaw in spatialInference as well as the filtered metricRes for comparison.
• implemented option to add different QC metrics of the functional model and plot them with different shapes in plotCrossHeatmap
• adapted the overview vignette to plot the raw spatial statistics curves
• accelerated all examples and tests with functional GAM via algorithm = "bam"

spatialFDA 1.1.15

• added a test to check the edf values and order from the RSE calculation. Made the RSE code more robust.
• added progressbar in crossSpatialInference.
• adapted the overview vignette to reflect the recent changes.

spatialFDA 1.1.14

• change the sum of mean residuals per condition to be the residual standard error. The main reason is to have a more comparable estimator of model fit by dividing by the degrees of freedom per condition. The degrees of freedom are the no. of data points per condition (no. of curves * data points per curve)

• the sum of the estimated degrees of freedom of the parameters per condition as provided by the summary.pffr() output.

spatialFDA 1.1.13

• provide the condition-wise sum of mean residuals across the curves. This is a measure to judge the quality of the fitted functional GAM by condition.
• allow for an automated threshold selection delta in spatialInference based on the mean over all images of the minimum nearest neighbour distance between points. Added test for this use case

spatialFDA 1.1.12

• changed the thresholding in calcMetricPerFov from having at least 1 point per (sub)point pattern to at least two in order to avoid having images with way to uncertain curves. In order to accommodate for this change some tests and examples had to be adapted
• spatialInference now prints the adjusted R squared to quickly assess overall model fit. As this is just one summary, still rigorous assessment via qq plots and autocorrelation should be performed.

spatialFDA 1.1.11

• Change in the weighting procedure in spatialInference and no calculation of weights in calcMetricPerFov. It is now possible to define custom weights, by equal weights, by the total number points, and for multitype processes by the smaller point pattern (min) and the larger point pattern (max)
• Option to extract the mean functional coefficient of the functional GAM over the domain $r$ as a measure of the effect size in extractCrossInferenceData
• plotCrossHeatmap creates now a bubble plot with the effect size being the colour and the size of the bubble being the p-value

spatialFDA 1.1.10

• p-value adjustment option in plotCrossHeatmap and filtering of coefficients for the heatmap. Fix some small global variable defintion errors

spatialFDA 1.1.9

• enabling Fisher's variance-stabilising transformation in spatialInference as it is recommended for $G$ functions.

spatialFDA 1.1.8

• adding a constant to the log in plotCrossHeatmap to avoid NULL values being plotted

spatialFDA 1.1.7

• Fixing error in plotCrossHeatmap when the model is NULL.

spatialFDA 1.1.6

• Error handling in spatialInference when one condition has no images with curves.
• Wrote as well a new test for this case

spatialFDA 1.1.5

• New overview vignette, showing the usability of spatialInference
• Plotting of a heatmap from extracted from crossSpatialInference
• Adding of the package sticker to the README

spatialFDA 1.1.4

• functionalGAM accepts flexible naming of the intercept column.

spatialFDA 1.1.3

• changed the default of the GAM estimation to be a Gaussian with log link for positivty of the response

spatialFDA 1.1.2

• wrote a new convenience function crossSpatialInference which makes the estimation across a range of cell types easier.

spatialFDA 1.1.1

• wrote a new convenience function spatialInference which makes the FDA estimation of the spatial statistics functions simpler
• in order to do this various changes to prepData were introduced
• furthermore, small syntactic changes for the retrieval of the functional intercept were added to functionalGam and plotMdl
• The vignette was simplified to reflect the changes in prepData

spatialFDA 0.99.14

• changed the example in the Vignette to scaled t-dist. with log link due to lower AIC than Gaussian.

spatialFDA 0.99.13

• changed the example in the vignette to Gaussian with log link and smooth random intercept.

spatialFDA 0.99.12

• Changed indexing of spatial coordinates to positional index. Names can differ between objects, e.g. "x" or "coord_x" etc. therefore, positional index generalises better.

spatialFDA 0.99.11

• major changes in fixing the levels of the mark factors in the point pattern. Prior to this fix, it could happen that the marks in selection do not correspond to the marks in the point pattern and therefore give the inverse result.

spatialFDA 0.99.10

• minor bug fix

spatialFDA 0.99.9

• remove imcdatasets dependency and instead download example data directly from ExperimentHub
• added small reader function .loadExample to download the data mentioned above

spatialFDA 0.99.8

• more packages explicitly named
• vignette shows use of PCA based random errors and exchanged one sample

spatialFDA 0.99.7

• explicit package naming for functions to make compatible with linux distributions
• possible to pass ... parameters to fPCA method

spatialFDA 0.99.6

• small error correction in show fPCA method

spatialFDA 0.99.5

• change in how failures in calcMetricPerFov are handled. Works now as well if rSeq = NULL

spatialFDA 0.99.4

• variable ID is no longer created in calcMetricPerFov
• adjusted examples and vignette to the above
• added option to extract gene expression directly in calcMetricPerFov if continuous = TRUE
• small bug fix in calcMetricPerFov to pass ... in all cases
• added variables legend.position and ncol to plotMetricPerFov

spatialFDA 0.99.3

• adjusted by variable to work with length 1
• adjusted ... in plotMetricPerFov to be passed only to geom_line

• adjust examples to build in under 10 min

spatialFDA 0.99.2

• Rewrote vignette with smaller dataset
• Implemented quasi likelihood family in vignette to improve variance estimation
• Adjustd calcMetricPerFov.R to be able to calculate cross functions

spatialFDA 0.99.1

• Adressed Bioconductor review
• Added unit tests
• Added functional boxplots
• Added diagnostic plots to vignette

spatialFDA 0.99.0

• Initial Bioconductor submission.