We offer external services on data/code management and statistical analyses. Your time can be better invested on you project's science, the generation of new ideas, and the writing of publications and proposals.
Examples of our services:
Data quality control: we can check your data for outliers, missing data, inconsistencies, etc. We write reproducible code for data cleaning and imputation customized to your needs.
LiDAR data processing. Generation of digital terrain models, digital surface models, and canopy height models. Growth estimates in tree height and crown size.
Image segmentation to semi-automate crown delineation using high spatial resolution imagery.
Spectral and spatial image analyses. Photogrammetry to estimate forest-level and tree-level properties.
Null models. Patterns are often influenced by system properties that must be controlled to avoid spurious conclusions, such as species diversity, aggregation of habitat or individuals within species, or differences in body size or abundances. We have developed a generalized approach to null modelling that adjusts hypotheses acceptance thresholds to constraint for Type I error (false positives) while minimizing Type II error (false negatives). This allows the "safe" testing of many hypotheses (e.g. associations between many pairs of species, or between many species with many environmental variables, etc.) without losing power.
Clustering and ordination of multidimensional data. From the standard tools in community ecology (PCA, PCoA, NMDS, PCNM) to more general and less used solutions for highly dimensional/disconnected data (curvilinear or non-linear component analysis, Sammon's mapping, LTSA, t-distributed stochastic neighbor embedding).
Variation partitioning analyses, both using dissimilarity modeling and redundancy analyses. We designed a way to estimate adjusted R-squared in matrix regressions. We improved the stepping stone algorithm so dissimilarity modeling conforms to the standard assumptions of linear modeling.
Network analyses and related applications. Many types of ecological data can be expressed as networks, opening the door to using many concepts and algorithms from graph theory. We have worked on the description of food web topologies, detecting cyclic succession through time, description of spatial associations between species, 'death spirals' in trees through the accumulation of damage, ...
We review cases for free, no strings attached: data@oikobit.com