Data Analytics
Any dataset that can be separated into independent variables and dependent variables can easily be analyzed on several levels. We can perform correlational analysis, principal component analysis, regression analysis (multiple regression, moderation, mediation, AN(C)OVA), and machine learning.
Mathematical models
A mathematical model makes use of known mathematical solutions to problems and concepts and uses them to create a mathematical representation of a system. Statistics inform our confidence about the model, given assumptions about our data sampling. Mathematical models are generally limited in scope, but often quicker and easier to create than other methods. In many fields of research, linear regression is a mathematical model commonly used to draw correlations between data.
Machine learning
Machine learning (ML) models can be specific or general in scope. Generally, ML models are data representations using mathematically proven and/or empirically supported optimization algorithms. Machine learning models often can be difficult to interpret because they are only custom-trained to specific data, but not custom-designed to specific data or theory.
Natural language processing (NLP)
- Generating a vectorized form of a document and/or individual words, which can be used for analysis as well as model inputs to train a model to predict something The vectorization involve topics/themes from LIWC, N-grams (groups of N consecutive word lemmas), and Word2Vec
- Generating GenSIM LDA topic analyses
- Performing sentiment analysis and entity extraction
For inquiries to conduct a specific algorithm from an open source library or one available to purchase, please email us with the specifics of your project and goals.
Data Dashboard and Visualization
In this kind of project, we apply likelihoods to model input distributions, and run the model with those probabilities to calculate most likely outcomes. Generally, the outcomes should be relatively simple, perhaps even simpler than the model output. The specificity of outcomes we allow ourselves to predict depends on the model’s level of validity. With a data set and a dashboard design, we can implement a dashboard or exploration tool that presents the data set as specified.