Parameter limits
Specify a lower and upper limit, with a comma in between. The limits will set the range on the x-axis in the various plots, and the range covered by the bins in the roulette method. The limits will also affect the fitted distributions: the gamma, log normal and log Student-t will all be shifted to have support over the range [lower, infinity), and the beta distributed will be shifted and scaled to have support over the range [lower, upper].
Elicitation method
For the General method, specify (any number of) cumulative probabilities P(X<=x).
- Specify values x1, x2,... in the Parameter values box, in increasing order, separated by commas.
- Specify the corresponding probabilities P(X<=x1), P(X<=x2),... in the Cumulative probabilities box, separated by commas. The requirements for each distribution to be fitted are as follows:
- Normal: smallest elicited probability less than 0.4; largest elicited probablity greater than 0.6.
- Student-t : smallest elicited probability less than 0.4; largest elicited probablity greater than 0.6.
- Gamma : finite lower limit and at least one elicited probability greater than 0 and less than 1.
- Log normal : finite lower limit; smallest elicited probability less than 0.4; largest elicited probablity greater than 0.6.
- Log Student-t : finite lower limit; smallest elicited probability less than 0.4; largest elicited probablity greater than 0.6.
- Beta : finite lower limit; finite upper limit; smallest elicited probability less than 0.4; largest elicited probablity greater than 0.6.
- Mirror gamma : finite upper limit and at least one elicited probability greater than 0 and less than 1.
- Mirror log normal : finite upper limit; smallest elicited probability less than 0.4; largest elicited probablity greater than 0.6.
- Mirror Student-t : finite upper limit; smallest elicited probability less than 0.4; largest elicited probablity greater than 0.6.
If the Roulette method is selected, click on the Roulette tab and click directly in the plot to allocate probs. Click just below the line at 0 on the y-axis to clear a column.
Output tabs
- PDF. A display of the fitted density function, if show fitted PDF is selected, using the selected Distribution. Fitted quantiles are displayed using the red shaded areas, if the quantiles are within the displayed range set by Parameter limits
- CDF. A display of the fitted distribution function, if show fitted CDF is selected, using the selected Distribution. Fitted quantiles are indicated, and the quantiles are within the displayed range set by Parameter limits
- Tertiles. Shows three equally likely regions, as specified by the tertiles, and the median (indicated by the dashed line). The tertiles displayed will either be the elicited tertiles, if they have been provided, or estimates obtained by linear interpolation of the elicited probabilities, with the Parameter limits assumed to have cumulative probabilities of 0 and 1.
- Quartiles. Shows four equally likely regions, as specified by the quartiles. The quartiles displayed will either be the elicited quartiles, if they have been provided, or estimates obtained by linear interpolation of the elicited probabilities, with the Parameter limits assumed to have cumulative probabilities of 0 and 1.
- Roulette. If the Roulette elicitation method is selected, this will display the roulette grid in which probability judgements can be specified.
- Compare group/RIO. If you have separately elicited probabilities from experts individually using the multiple experts app, and you are using this app to obtain the consensus distribution (the "Rational Impartial Observer's" distribution), you can compare the result in this tab. In the multiple experts app, you need to download the judgements as a .csv file. You can upload the .csv file here, and then see a comparison between the RIO distribution, the experts' individual distributions, and a linear pool of the experts' distributions.
Download report
This will generate a pdf file, with each fitted distribution specified.