892 of the 4500 bank accounts have credit cards. There are three classes of card, 'junior', 'classic', and 'gold'. Here we ask the question, "Does the average account balance differ based on the credit card class?"

The following graphs are percentile boxplots (100th, 75th, 50th, 25th, and 0th percentiles, plus outliers) of the average account balance for each account with a credit card. The two plots show the same data, but grouped by synthesis method on the left, and by card type on the right.

{% include image.html src="/assets/img/trans_account_card_clients.avg_balance.card_type.png" alt="SynDiffix usage style" max_width="650px" %}

The SynDiffix synthetic data was produced by first computing the average balance per account from the original data to produce a two-column table consisting of the `avg_balance` and the `card_type`, with one row per account (i.e. not time-series data). This two-column table was then synthesized.

From the original data, we see that indeed gold credit card holders have higher average account balances than classic card holders, which in turn are a little higher than junior cord holders. The data for SynDiffix is very close to the original data, and accurately shows the same trend.

The data for Mostly AI and CTGAN are less accurate. Both show a slightly higher median balance for gold cards, but fail to show a lower balance for junior cards. Both Mostly AI and CTGAN show a much wider range of average balances, include negative average balances where none should exist.