Web > Charts > Chart Type Walkthroughs > Cartesian

Boxplot of Spreads by Venue

Known issues

Some chart features have known issues in the current release. See Known issues.

This walkthrough draws a box and whiskers per venue from 5 precomputed values, answering "Which venue has the widest spreads?"

Prerequisites

• Editor mode: click once on the screen, then press Ctrl+d. The AMIDB Shell Tool and the gallery are used as in the quickstart.
• Data: the example data that stage 1 creates. To use a table of your own instead, see With your own data.

1. Create the example data

The chart reads a table in the Center, so create the table first.

1. Select Dashboard -> AMIDB Shell Tool....

2. Run:

   CREATE PUBLIC TABLE ChartDemoSpread (Venue String, MinBp Double, Q1Bp Double, MedianBp Double, Q3Bp Double, MaxBp Double);
   INSERT INTO ChartDemoSpread VALUES ("NYSE", 0.8, 1.4, 1.9, 2.6, 4.1), ("NASDAQ", 0.6, 1.1, 1.6, 2.2, 3.5),
    ("BATS", 1.0, 1.8, 2.4, 3.3, 5.2), ("IEX", 0.9, 1.5, 2.0, 2.7, 3.9);
   SELECT * FROM ChartDemoSpread;
   

3. Check that the lower pane shows the 4 rows.

The template draws the 5 values it is given and computes nothing, so the table already holds each venue's minimum, quartiles, median, and maximum.

2. Create the chart

The gallery creates a datamodel that reads the table, then places the chart.

1. Select Windows -> New Window, then the edit button on Blank1, then "Create Table / Visualization / Form".
2. Select "AMI (__AMI)", then "Next >>".
3. Select the "USER" tile ChartDemoSpread, then "Next >>".
4. Select "Test (alt+enter)", then "Next >>".
5. Select the "Cartesian" tab, then the "Boxplot" tile.

6. Set the fields below the tiles:

   Field	Value
   "Category:"	Venue
   "Min:"	MinBp
   "Q1:"	Q1Bp
   "Median:"	MedianBp
   "Q3:"	Q3Bp
   "Max:"	MaxBp

7. Select "Finish".

8. Check that the window shows 4 boxes, BATS, IEX, NASDAQ, and NYSE, each with a median line and whiskers.

Result

4 boxes. BATS has the highest median, 2.4, and the longest whisker, to 5.2; NASDAQ the lowest, 1.6.

The chart's own PNG export, from AmiScript toImagePng; see Chart type walkthroughs.

To check without looking, read the render stats: session.getDevTools().getAllRenderStats() in the AMIDB Shell Tool, with "Run on this web session" selected (see Running a script), or web_debug op=panelRenderStats over MCP. The "Box", "Median", "Whisker", "Min Cap", and "Max Cap" layers each report Rows 4 and an empty Problem.

With your own data

If your table holds one row per observation rather than the 5 values per category, shape the rows in the datamodel's query. To try it on example rows, run this in the AMIDB Shell Tool first:

CREATE PUBLIC TABLE ChartDemoFills (Venue String, SpreadBp Double);
INSERT INTO ChartDemoFills VALUES ("NYSE", 0.8), ("NYSE", 1.4), ("NYSE", 1.9), ("NYSE", 2.6), ("NYSE", 4.1),
 ("NASDAQ", 0.6), ("NASDAQ", 1.1), ("NASDAQ", 1.6), ("NASDAQ", 2.2), ("NASDAQ", 3.5),
 ("BATS", 1.0), ("BATS", 1.8), ("BATS", 2.4), ("BATS", 3.3), ("BATS", 5.2),
 ("IEX", 0.9), ("IEX", 1.5), ("IEX", 2.0), ("IEX", 2.7), ("IEX", 3.9);

Remove it afterwards with DROP TABLE ChartDemoFills;.

1. Follow stage 2 with the "USER" tile ChartDemoFills, or the one for your own table, in place of ChartDemoSpread.

2. When the gallery shows the datamodel's query, replace the query with:

   {
     CREATE TABLE ChartDemoSpread AS EXECUTE SELECT Venue, min(SpreadBp) AS MinBp, percentileCont(SpreadBp, 0.25) AS Q1Bp, median(SpreadBp) AS MedianBp, percentileCont(SpreadBp, 0.75) AS Q3Bp, max(SpreadBp) AS MaxBp FROM ChartDemoFills WHERE ${WHERE} GROUP BY Venue;
   }
   

3. Select "Test (alt+enter)".

4. Check that the "Tables" tab shows ChartDemoSpread alone, with the same 4 rows as the example table in stage 1.

Then continue with stage 2: the fields are the same, and any later stages apply as written. To do this over MCP, put the same statements in the onProcess of the datamodel that step 1 of "Over MCP" imports, with USE ds="AMI" before each EXECUTE.

percentileCont interpolates between observations; percentileDisc takes the nearest one instead.

Over MCP

An agent builds the same chart from the same table. Connecting an agent is covered in the AI section; the tools are on Building charts over MCP.

1. Create the datamodel with web_execute op=importDatamodel and this dmJson, then run it once with op=initDatamodelSchema and dmName ChartDemoSpread:

   {"id": "ChartDemoSpread", "queryMode": "startup", "callbacks": [{"name": "onProcess", "amiscript": ["{ CREATE TABLE ChartDemoSpread AS USE ds=\"AMI\" EXECUTE SELECT * FROM ChartDemoSpread; }"]}]}
   

2. Create the chart:

   {"windowName": "ChartDemoBoxplot", "templateKey": "tplBoxplot", "datamodelId": "ChartDemoSpread", "dmTable": "ChartDemoSpread", "title": "ChartDemoBoxplot", "bindings": "{\"CAT\": \"Venue\", \"min\": \"MinBp\", \"q1\": \"Q1Bp\", \"median\": \"MedianBp\", \"q3\": \"Q3Bp\", \"max\": \"MaxBp\"}"}
   

3. Run the datamodel again with web_execute op=executeDatamodel, dmName ChartDemoSpread. A chart created over MCP can stay empty until its datamodel runs again, with its axes at 0-100.

4. Confirm that every layer reports the rows above and an empty Problem: run web_debug op=panelRenderStats.

The call is web_create_chart_from_template.

Removing the example

Remove the chart, then its datamodel, then the table, so that the next walkthrough can reuse the names:

1. Close the chart's window with its close button, and select "Delete" when asked whether to delete the panel.
2. Select Dashboard -> Data Modeler..., expand "Datamodels", right-click the chart's datamodel, ChartDemoSpread, and select "Delete Datamodel", then "Yes". If that name was taken when stage 2 ran, the gallery added a number to it, as in ChartDemoSpread1.
3. Run DROP TABLE ChartDemoSpread; in the AMIDB Shell Tool.
4. Check that SELECT * FROM ChartDemoSpread; reports Unknown table: ChartDemoSpread.

Over MCP, delete the chart with web_danger op=deletePanel, then the datamodel with op=deleteDatamodel, dmName ChartDemoSpread, and drop the table as in step 3. deleteDatamodel refuses a datamodel that has been committed; delete that one in the Data Modeler as in step 2. See Removing an agent's chart.