Quantitative Data Analysis

People who say that statistics don't matter usually don't understand them. 

Understanding statistics is crucial in designing effective baseline and monitoring programs that are actually capable of signalling when a change in an environmental parameter has occurred.  With a poor understanding of statistics, study design, statistical power, and the ultimate tests that will be used in the final analysis, an environmental parameter being measured may have to deviate an unacceptable magnitude from poorly measured background or target levels in order to detect a statistical change. This fact no longer flies over the heads of regulators, who will ask that you prove that your study approach and data set can actually detect differences statistically (i.e., they will ask you for the statistical power). Frankly, without statistics integrated into your project from day one, you could be wasting years trying to answer questions that can't be answered, or which could have been addressed in a shorter period of time, using more direct, and correct, approaches. 

We understand statistics, even the really messy stuff. We integrate statistical due diligence into all of our quantitative studies, from start to finish. 

Software Used 

We are capable of using over five statistical programs, including SPSS, SYSTAT, MVSP, JMP, G-POWER, and R, but primarily use R or SPSS 22.0 Professional with advanced add-on packages, including those for mixed models, and complicated multivariate procedures. SPSS 22.0 Professional provides the user with the ability to interface with R-Script, and so we are also able to integrate or create R script, and even override assumptions that are not appropriate for a particular situation (for example, when adjacency effects are present). We have overwhelmingly found that this combination of programs allows us to do the same analyses as those struggling with R or other programs alone in less than half the time, and with greater accuracy. 

Qualitative Data Analysis

Not all data are meant to be analyzed using statistics. Examples of such data include traditional knowledge and personal observations, as uncovered through interviews. Our in-house Aboriginal engagement specialist is a highly experienced social sciences researcher, who employs the latest qualitative research techniques, and programs such as Nvivo, which allows the practitioner to identify trends in values, concerns, attitudes, and other nodes of interest. You'd be surprised at how much can be learned from interviews with knowledgeable people, and we have the right tools and experience to extract and present that information in a very usable manner. 

SAMPLE STATISTICAL PROJECTS

Sample Project 1: Analysis of energetic and movement responses of barren ground caribou to tundra roads, NWT

The EKATI Diamond Mine (EKATI), owned and operated by BHP Billiton Diamonds Inc., is located in the Northwest Territories, approximately 300 km northeast of Yellowknife. For the Ekati project, BHP constructed several all-season haul roads made from gravel crush, with berms of varying heights and rock composition. In 2011, there was a renewed interest in determining whether the road structures and traffic could impact caribou safety and movement patterns. To help address this concern, Zoetica was contracted to analyze motion triggered and timed camera monitoring data to determine how roads were impacting caribou behaviour and movement patterns.  Zoetica used the GENMOD procedure in SAS 9.3 for most comparative analyses, which fits generalized linear models. For statistical comparisons of caribou encounter rates, a Poisson distribution was used with a log link function and an offset term to account for camera effort. Poisson regression analyses are appropriate for modeling count and rate data. For statistical tests comparing behaviours that were or were not observed (P/A), logistic (binomial) models with a log link function were used. Type 3 Likelihood Ratio (LR) Statistics were most often requested from the GENMOD procedure, which are similar to the Type III sums of squares in PROC GLM, except that likelihood ratios are used instead of sums of squares, and maximum likelihood estimates are computed under the constraint that the Type III function of the parameters is equal to 0, by using constrained optimization. Variations of these tests were done as appropriate, and where statistical power was sufficient. Overall, 11,112 behaviours were recorded in 6,399 individual caribou, which allowed for general conclusions to be reached regarding time periods where caribou movement across roads was most prevalent, specific effects of roads and group size on crossing behaviours, crossing rates among locations, and primary causes of caribou deflections from roads. Conclusions allowed for spatially and temporally appropriate mitigation. 

Sample Project 2: Effect of Hydro Power on Invertebrate Populations and Abundance

Data Zoetica was contracted to analyse Invertebrate data from a major run-of-river project in BC. Data were collected between 2002 to 2008, to determine whether or not the hydro-electrical project was impacting local invertebrate communities, which are a major food source for fish populations. Zoetica analyzed benthic samples collected using drift net and surber sampling techniques separately in SPSS, using advanced statistics package bundle. Both data sets were analyzed using mixed models, which were ranked using AIC criteria. General results were confirmed by application of a non-parametric Friedman’s 2-way ANOVA on ranks. These two tests are appropriate where a repeated measures sampling design has missing data for some years, or low sampling effort, both of which applied to this data set. Data collected using drift net sampling and surber sampling suggested that site played a near-significant (according to drift net samples) or significant (according to surber samples) role in predicting invertebrate numbers, with generally higher numbers at the intake location compared to the powerhouse location. The surber data set, which had a greater statistical power and more years of sampling, allowed for the detection of 4 significant effects in the prediction of invertebrate numbers: Site, Year, Genus, and a 3- way interaction (Site*Year*Genus). The 3-way interaction suggested that the invertebrates at the intake and powerhouse locations reacted differently, likely in response to the introduction of the project, and that Diptoptera and Ephemeroptera reacted differently than Tricoptera and Plecoptera. The effect was most pronounced on Ephemeroptera, which increased substantially in 2004, but then returned to pre-commissioning levels by 2005. The increase seen in Diptopera in 2005 was less pronounced, but numbers returned to pre-commissioning levels between 2006 and 2007. 

Sample Project 3: Effect of Coastal Gillnet Fisheries on Seabird Net Entanglement Risk 

Zoetica was contracted by Environment Canada to analyse data from an observational study of seabird abundance and behaviour around commercial gillnet fishing vessels to evaluate net entanglement risk in coastal BC. Our analyses focused on rhinoceros auklets (Cerorhinca monocerata), the most abundant seabird detected during at-sea transects. Auklets are epipelagic foragers, diving for long periods to retrieve fish, including juvenile salmon (Onchorhyncus spp.), which puts them at risk of becoming entangled in salmon gillnets. As seabird by-catch is suspected to be under-reported in the study area, we used indirect measures and statistical inference techniques to evaluate risk. Auklet counts and behaviours were recorded during at-sea transects in commercial fishing areas and point count surveys at active gillnet vessels. Advanced zero-inflated mixture modeling techniques were used to estimate probability that diving birds were present during surveys, but missed during counts. This analysis revealed a higher probability of unrecorded diving birds closer to fishing vessels. Results of the study suggested that patterns of relative auklet effort-adjusted density and behaviour change near gillnet fishing vessels that could suggest attraction-related patterns of risk. The results, however, also showed that site, time of day, tide height, and distance from shore were important environmental predictors of auklet densities, while net depth and mesh size were significant parameters that predicted auklet densities near gillnets. Results have important implications for mitigating potential impacts of coastal gillnet fisheries on this species and other alcids with similar foraging traits. 

Sample Project 4:  Power Analysis of Amphibian Monitoring Program (Parks Canada), Alberta

 Analyzed monitoring data to determine whether Parks Canada’s amphibian monitoring program could detect a 30% or greater decline in amphibian populations in Waterton National Park over the next ten years. A power analysis was conducted using occupancy models, which showed that a monitoring plan redesign, utilizing more sampling sites, was required. A report of statistical results and detailed recommendations was produced. This project was completed by Dr. Bears, Ms. Rogers, and a ZoeticaNet partner from the actuarial department at Simon Fraser University.