How to use the estimator module
The Metrisim estimator module estimates partworth utilities, for every survey respondent. In essence, partworths shows the impact of each attribute level on a survey respondents overall rating of the product profile. The simulator sums the partworths relevant to each product in the scenario to produce a utility score and then uses a decision rule to translate this into either a prediction of which product a respondent will choose (Max Utility decision rule) or a probability of purchase (BTL). Further adjustments may occur if you use Metrisim's relevant set indicators or calibration. All respondent predictions are then used to produce a share % which you see in the second to last row of the control panel of the simulator.
The impact is first estimated as a regression coefficient and then mean centered to produce partworths. This is for ease of interpretation when viewing the partworth average plots.
Note: if you are an R conjoint package user you may get slightly different results when saving to a csv. This means that if you upload the partworths from the R conjoint package to the Metrisim simulator, it will also give slightly different results when running scenarios. This is because Metrisim uses higher precision (more decimal places) while R conjoint - at least in terms of what is saved to csv, if not internal calculations - uses less decimal places.
Steps:
1. Upload ratings.csv
Ensure you have ratings from your survey and you have uploaded this as a ratings.csv. The number of rating columns must equal the number of product profiles (i.e. rows / runs in your experimental design). If you want you can leave any id column in, it will be ignored.
Example of ratings.csv layout
2. Upload attribute_map.csv
If it is no longer present, as you overwrote it when switching to another project, reupload.
Example of attribute_map.csv
3. Upload design_text.csv
If it is no longer present, as you overwrote it when switching to another project, reupload.
Example of design_text.csv
4. Click 'Run'
Click run. Once it has finished processing, you can also download a zipfile containing the partworths (pws.csv) and the R square per respondent. Click download then wait for the link to appear and click on that to download.
Estimator module screenshot
5. Respondent quality screening
If the R square is below a threshold you determine, such as 0.4, you might exclude such respondents from the ratings file and then reupload and repeat the above steps again. Alternatively you could delete respondents in the pws.csv you downloaded, and then reupload for the simulator to use - however - if you accidentally ran the estimator again it would overwrite pws.csv and you would include any sub-standard respondents in the new pws.csv as they were in the ratings.csv which is used as input into the process.
Metrisim experimental design generator screenshot
Example
Copyright reserved, Craig Kolb, 2025