## Week 14: Survey Data: Weighting | DS23

by Auguste Navickyte

When using survey data, you may want to look at weighted numbers to avoid a biased analysis. Below, I summarise how your workflow should look like - whether in Tableau Prep or Alteryx Desktop.

### Data Fields

Fields of interest should come from two tables: our initial survey and population data.

Example fields from initial survey:

• Respondent ID
• Country
• Activity (or any other measure)
• Number of respondents within each country and activity (you can get that by calculating countd(respondent ID))

Example fields from population data:

• Country
• Activity (or any other measure)
• Population within each country and activity

### Steps to follow

1. Find the total number of respondents
2. Find the share of respondents within each country and activity: A = countd(respondent ID) / no. of respondents
3. Find the total number of all populations
4. Find the share of each population within all populations: B = population / all populations
5. Calculate weight multiplier for each unique group*: B / A
Note: the weight multiplier for the underrepresented group should be >1 whilst the opposite is true for the overrepresented group
* In this example, unique groups are based on country and activity
6. Adjust results (in this case, the number of people taking part in certain activities) based on the weight multiplier