From the "Add Item" menu select “Blanks”
The Context and Stem
In “Context” you should write the context and stem
You can use markdown to write the context and stem
Names of data, functions, columns, numeric values etc. should use code format
Never use bold/underline/italic
Use R/Python output to format short tables (no more than 5 lines)
Use schema diagrams to explain SQL data structures
Start a new sentence on a new line, with a blank line between (see example below)
Pre-challenge code
Include any code that is needed to set up your item, but does not need to be seen by the test taker.
Importing/creating data
Defining the database (SQL)
Pre-cleaning
Model fitting (e.g. for modeling items focused on metrics)
Code 1
This is what the user will see.
Keep this short and related to the point being tested
Only include code that is relevant (e.g. include package loading but don't include data prep)
Use tags where you want to leave a space for the user to complete. The tags need to follow the pattern:
|
We use expr1, expr2 etc. in the tags so we don't accidentally use reserved words and break the build
You need to follow the tag pattern exactly for the blank to appear correctly
You can have more than one blank in an item
Variables
This is where we define what the correct answer is.
Use the label inside each tag to identify the blank
Leave a space before and after the bullet "-"
Write the answer inside single quotes
Use double quotes to define strings in the code
You won't be able to preview your item until you define what goes in the blank
Before saving:
Add a descriptive title - summarise what you are testing e.g. "Calculate probability of an event"
Add your Item Writer ID (we will tell you what your ID is)
Preview the item to make sure it looks as expected
Common Problems
If the preview or build fails there are a few common issues that may be the cause.
Have you defined a variable that will be the blank in the code?
Does the code run in your local environment? If it does, what is the difference?
Have you missed out loading a package?
Is the data the same?