Description
Polling is subject to various factors that distort the usefulness of the results. Certain groups respond more frequently than others, biasing the overall tally towards their views. Weighted polling can compensate for this effect.
Proposal
Add an option to weigh poll results by demographics.
Concept
Most professional polling results are weighted using demographic information. The respondents are divided into cells, usually by age, gender, income, race/ethnicity, etc. The individual responses are then weighted by the respondents representation of the general population.
Example
For example, if your website has the following population:
| Women | Men | |
|---|---|---|
| Below 35 | 45% | 25% |
| 35 and Above | 20% | 10% |
and your current polling had the following average rating per cell:
| Women | Men | |
|---|---|---|
| Below 35 | 2.34 | 3.54 |
| 35 and Above | 1.96 | 4.32 |
The weighted rating would be:
2.76 = (.45) * (2.34) + (.20) * (1.96) + (.25) * 3.54 + (.10) * 4.32
Methodology
The demographic data for individuals could be collected using the profile module. The weighted polls could then be calculated with a custom calculation as described in the API document.