When creating new items in multi-level projects, it's important to consider the data level you are working in. When you create the new construction you need to specify the unit of count. You can create a construction in a level using the same unit of count, or you may want to specify a different unit of count.
In this article, we will look at a multi-level project with respondents at the parent level and occasions at the child level. Respondents have been asked about beverages they have consumed across different occasions, so one respondent can have multiple occasions.
It is important to understand the way the different levels of counting work in multi-level projects and if you are not familiar with multi-level projects and units of count, please review Multi-Level Data and Multi-Level Analysis.
In this article
1. Specify the Unit of Count
Respondent Level
When creating a new item in a multi-level project, Harmoni specifies the unit of count and defaults to the parent level label.
- Add the new item and the unit of count label appears next to the axis name field.
- Add the axis name and press enter.
- You can now edit the unit of count label. In this example, we will leave it as Respondents by pressing enter.
- Enter the element labels.
- Click new axis
- Enter the axis name
- Press enter in the unit of count level
- Enter the element name
- Define the new element
Learn more about creating your own constructions.
Occasion Level
When creating a construction at the child level, Harmoni defaults to the parent level label (Respondents) and to build your construction using the occasion level, you need to enter Occasions as the unit of count.
Here we will create an occasion level construction using Weekday and Morning.
- Click new axis
- Enter the axis name
- Enter Occasions for the unit of count
- Enter the element name
- Define the new element
Review the labels in the COUNTING section at the top of the Project tree to ensure you spell the unit of count label correctly.
If you misspell the word, Harmoni provides a warning and your construction is not retained.
2. Unit of Count Calculation
Sometimes, you may want to see the occasion data at the respondent level. For example, analyzing brand occasion level data, at the respondent level, can allow you to understand for drinkers of Brand A, what other brands they are drinking. This type of analysis is not possible at the occasion level.
In the example below, the brand axis taken to the respondent level can be crossed by itself to reveal who else is drinking each brand.
Occasion Level Data Occasion data at the Respondent Level
Changing the unit of count is as simple as entering the required unit of count when creating the new item. Using an example with Weekday/Weekend and Daypart, the image below shows the Occasion level construction and the new Respondent level construction. The axis labels are also updated to include the data level for clarity.
When you create new items using the different levels of count, it is important to note that you may see differing results despite the construction definitions being the same. The way Harmoni counts the data are as follows:
- When you create a new item in the occasion data and you specify occasions in the construction, Harmoni looks at the occasion records.
- When you create a new item in the occasion data and specify respondents in the construction, Harmoni looks at the respondents who had occurrences of these occasions.
Counting respondents versus counting occasions can lead to different results
To explain this clearly, we will use a simple example with five respondents and eight occasions, and we will construct a new item to show drinkers of brand A in the morning (Brand Consumed=A AND Day Part=Morning). The data is occasion level, but we will specify occasions in one construction and respondents in the other.
The constructions are defined like this:
If you construct the item and define it using Occasions, Harmoni will count the occasion records of Brand Consumed=A and Day Part=Morning.
Running an analysis will give you:
- 2 Respondents - IDs 1 and 3
- 2 Occasions - IDs 1.1 and 3.1
If you construct the item and define it using the respondent level, Harmoni counts the records of respondents for whom Brand Consumed=A and for whom Day part=Morning
All respondents who consumed Brand A
All respondents who have a morning Day Part
Running an analysis gives you:
- 3 Respondents - 1, 3, and 4
- 5 Occasions - 1.1, 1.2, 3.1, 4.1, and 4.2
You can see that although respondent 4 consumes brand B, because it was consumed in the morning, they are included in the calculation. Likewise respondent 1 and 4 consume in the afternoon, but it's for brand A, and so they are included. All 5 occasions for respondents, 1, 3, and 4 are counted.
Where to from here?
Learn more about Harmoni