The App

The App Interface

The front end of the app has been design to be minimalist but with enough information available to the user to work out what’s going on. Figure 1 below shows the main area of the app that’s visible on loading.

Figure 1. The WhatsMatching app - main interface

When the app loads, the first simulation is run. This means that as soon as a user opens the app they are presented with a standard presentation of matched data to investigate. The full home screen of the app also contains a sidebar panel on the right with some additional information for users. Table 1 below outlines some of the options that users have to access through the app.

The four buttons under the main title each serve a different function outlined in the table below.

Table 1. The WhatsMatching app - primary user options
	Simulate the data and select the matching settings
	Look at the data and get insights
	Generate data from a random simulation with random matching settings
	Links to information about the app (this website)

Data Simulation

Selecting the icon opens up to options for data generation and settings for matching. There is further information available about the options available here vignette("DataGen", "WhatsMatching").

Data and Insights

Selecting the icon presents modal display with various information about the data that has been generated. This includes the same plots that are in the Data Generationsection. There is also a table of the data available; information about the match settings (also visible on the main page); insights about setting a calliper; and insights into the estimates of all methods with other regression model settings.

The calliper plots are similar to those laid out in vignette("Background", "WhatsMatching"). There are four plots that show the distances between matched pairs in an unadjusted format (top row) and a standardised format (bottom row). The standardised plots are measured in the number of standard deviations above zero. The plots on the left show the distances in the order they were matched while the plots on the right show them in order from closest to the furthest distance. If you want to apply a calliper to the matched data, find the observation that you think is the beginning of the outlier group you’d like to remove. Subtract one from that number and go to that frame using the slider on the main screen.

The plot of the estimates is similar to one shown in the Functions vignette. The user has the option of specifying which method they would like to see the estimates for. When Apply is click, the plot will generate the estimates with the 95% confidence intervals for all possible regression formulas for those methods for easy comparison.

Random Mode

Selecting the icon runs a random simulation with a random combination of match settings. If the user doesn’t have any ideas about what to investigate then they can push this button and the app with decide. It has been designed so that Method 1 and Method 2 will always be different.

Information Links

The opens this website to the homepage.

The `plotly` interface

The main app feature, the frame-by-frame visualisation of matching algorithms, has been generated using plotly. The plotly interface has some very nice and intuitive options for navigating visual data. Users are able to easily zoom in to plot areas and reset them using the buttons available on the top left of the plot. The items in the legend can be toggled out of the plot area to make visualising things easier. The slider paused or reset, allowing users to examine different points in the matching algorithm.