• 5areas: data, view, model, result, integrate
  • 3equation types: single, standard panel, dynamic panel
  • Desktopapplication for analysts
  • C#WPF, .NET Core and SQL

The challenge

Analysts need one dependable place to explore data, build models and share results.

Analysing data and building statistical models involves many linked steps. Data has to be loaded and cleaned, summarised across time periods, charted, transformed, modelled in several ways and then presented. Analysts also need to try many variable combinations and equation types, compare the results and reuse what already works.

The client had been using an earlier application built on VB. It produced errors at every step, which made the daily work slow and the results hard to trust.

The client needed one dependable desktop application where an analyst could take data from a CSV file, explore it, build models, review the results and share them, all in one place.

What we built

Catabatic worked with the client to customise the tool into a business intelligence desktop application. It is installed on the analyst's computer, and the work is organised around projects. Each project moves through five areas: data, view, model, result and integrate.

Data

Load, edit, summarise and label the data.

  • Start page

    Create a new project or a new update, activate a project, or delete one.

  • Edit data

    View and update uploaded CSV data. Copy, insert and delete columns, rename headers, and save the project, save it as a new project or save it as a new update.

  • Summary

    Summarised data by character and numeric variables, across yearly, half-yearly, quarterly, monthly, weekly, daily and total period views, depending on the type of data.

  • Report

    Apply labels to the data and export what is needed.

  • Pooling criteria

    Choose the column used for pooling. The ID column is selected by default.

View: charts and transformations

Explore the data visually.

  • Charts

    Drag variables to the x-axis and y-axis and the chart is drawn. Format the axes, rename series, change the chart type, apply display formulas such as normalise, percentage, average and split by time, and export the chart data to Excel.

  • Slide show

    Add charts to a slide show and view them together. A saved chart can be updated and saved again.

  • Chart transformations

    Draw charts for combinations of cross-section and variable, and apply several transformations to them.

  • Favourite transformations

    Build transformations once, save them and reuse them in chart transformations.

  • Analyze graph

    Apply lags to variables. The correlation and concordance values fill in a table, with concordant, de-concordant and transient figures.

Model

Single equation, standard panel and dynamic panel models.

  • Equation types

    Single equation, standard panel and dynamic panel. Each has a time component page for the model type, fit order, time series component, holdout period and model period.

  • Set equation

    Build equations from variable and cross-section combinations, apply transformations and run the model. One page serves every equation type.

  • Pooling

    Pool, un-pool or exclude a variable and cross-section in the pooling structure. Pooling control shows beta and t-statistic, Bayesian prior and lambda for dynamic panel models.

  • Output

    Output log, beta graph (line, clustered column and bell curve charts), model summary for every equation type, and the decomposition table with a real scale option.

  • Decomposition settings

    Settings applied when the decomposition runs.

  • Copy model

    Copy the time component, equation, decomposition parameters and variable settings between equation types, model numbers and cross-sections.

Result, sharing and settings

Review results and share them.

  • Model fit

    Chart of actual, predicted and error values.

  • Model decomposition

    Decomposition chart and table.

  • Drivers of change

    Waterfall chart and table.

  • Support and effectiveness

    Column chart comparing raw data with output data effectiveness.

  • Variable settings

    Aggregation structure, aggregation rule, and variable labels and groups for the results.

  • Share and get models

    Share a model and its settings with other users, and get models that others have shared.

  • Tools and themes

    Application settings, and themes to change how the application looks.

How it works day to day

  1. 1

    Start a project

    The analyst creates a project from the start page and loads the CSV data. Updates can be saved as new versions of the project.

  2. 2

    Prepare the data

    Columns are edited, summaries are checked by time period, labels are applied and the pooling column is chosen.

  3. 3

    Explore

    Charts are drawn by dragging variables. Transformations are applied and saved as favourites. Lags are tested to see correlation and concordance.

  4. 4

    Model

    The time component settings are saved, equations are built, and the model is run as a single equation, standard panel or dynamic panel model. Pooling is adjusted and the output is reviewed.

  5. 5

    Review results

    Model fit, decomposition, drivers of change and support and effectiveness are shown as charts and tables.

  6. 6

    Present and share

    Charts go into a slide show, data is exported to Excel, and models are shared with other users.

In use

  • What it is

    A desktop application for data analysis, modelling and reporting, installed with an installer package.

  • Who uses it

    Analysts, working in projects.

  • Technology

    Built with C#, WPF and MVVM, with DevExpress controls, on .NET Core with MVC and Web API, and SQL and SQLite databases.

Need a desktop application built?

Tell us what your users do every day. We will reply with how we would build it.

Contact us