KerusCloud® Parameter Converter
User Guide
Section titled “User Guide”This guide documents the KerusCloud® Parameter Converter (KPC) and explains how KPC fits into the KerusCloud® clinical trial simulation workflow.
Purpose and Audience
Section titled “Purpose and Audience”KPC converts distribution parameterization types commonly used in clinical trial simulations (e.g., Beta, Log-Normal, Weibull, Poisson).
This guide is intended for clinical statisticians and clinical trial modelers using KerusCloud®.
KPC matters because reliable parameter conversion reduces the risk of incorrect simulation assumptions and helps clinical teams translate statistical inputs into consistent KerusCloud® model parameters.
Quick Start — Using KPC
Section titled “Quick Start — Using KPC”1. Select the Distribution Type
Section titled “1. Select the Distribution Type”Choose from the Variable Type dropdown.
The interface currently supports:
- Beta
- Exponential
- Log-Normal
- Uniform
- Weibull
- Binomial
- Negative Binomial
- Poisson
2. Select the Method
Section titled “2. Select the Method”Pick the form that matches your available inputs (e.g., Mean & SD, Alpha & Beta).
3. Enter Numeric Values
Section titled “3. Enter Numeric Values”Fill out the input fields.
- Inline validation will show an error if a field is empty or non-numeric.
- Changing any input automatically clears previously computed results.
4. Click Convert Parameters
Section titled “4. Click Convert Parameters”Results appear in the results panel.
Use the Copy button next to each result to copy values for use in KerusCloud®.
Supported Distributions and Typical Clinical Uses
Section titled “Supported Distributions and Typical Clinical Uses”For all variable types the Mean and Standard Deviation parameters correspond directly to the mean and standard deviation of the distribution and hence are expected values for these summary statistics in the simulated data (virtual population). The converter supports the following distribution types and alternative parameters:
Convert to Shape1/Shape2 parameters from:
- Alpha/Beta (alternative names for conventional “shape” parameters)
- Mean/Sum (where Sum is the sum of the conventional shape parameters, and can be referred to as the “sample-size” of the distribution, particularly in a Bayesian interpretation)
- Mean/Standard Deviation
Exponential
Section titled “Exponential”Convert to a Mean from a Rate parameter (if the variable represents the conventional time to event of a process which occurs at a constant average Rate per unit time).
Log-Normal
Section titled “Log-Normal”Convert to Location/Scale parameters (where Location is the expected value on the natural log-scale and Scale is the standard deviation on the natural log-scale) from:
- Mean/Standard Deviation (on the raw/untransformed data scale)
- Median/Multiplicative Standard Deviation (where Median is the median on the untransformed scale which is equal to the exponential of the expected value on the natural log-scale, and Multiplicative Standard Deviation is the exponential of the standard deviation on the natural log-scale)
Uniform
Section titled “Uniform”Convert to Minimum/Width parameters (where minimum is the minimum of the distribution and Width is the width of the distribution) from:
- Minimum/Maximum
- Mean/Standard Deviation
Weibull
Section titled “Weibull”Convert to Shape/Scale (where these are the conventional “shape” and “scale” parameters) from:
- Shape/Median (where Shape is the conventional parameter and Median is the median, which may be useful if the variable represents a time to event process)
- Mean/Standard Deviation
Commonly used for:
- Time-to-event simulations
- Survival analyses
- Dropout simulations
Binomial
Section titled “Binomial”Convert to a Probability parameter (where the variable represents a single experiment with two possible outcomes, also known as a Bernoulli trial, and Probability is the probability of an outcome of 1 rather than 0) from a Mean.
Negative Binomial
Section titled “Negative Binomial”Convert to Event Probability/Number of Events parameters (where conventionally the variable represents the number of non-events which occur in a series of Bernoulli trials before the specified Number of Events occur, and the Event Probability represents the probability of the event occurring in each trial) from:
- Mean/Heterogeneity (where Heterogeneity is the formulation of the conventional “dispersion” parameter where values close to 0 correspond to lower variance and increasing values correspond to greater variance and “over-dispersion” relative to a Poisson distribution with the same mean)
- Mean/Standard Deviation
- Mean/Variance (where Variance is the variance)
Poisson
Section titled “Poisson”Convert to a Mean from a Rate (if the variable represents the conventional number of events occurring in a fixed unit of time when events occur at a constant average Rate per unit time).
Important Technical Notes
Section titled “Important Technical Notes”Weibull Mean/SD Conversion
Section titled “Weibull Mean/SD Conversion”Implemented using an iterative bisection search to determine the shape parameter:
- Computes the target coefficient of variation as
(standard deviation / mean)². - Uses gamma functions to compute distribution moments.
- Expands search bounds dynamically when needed.
- Runs a bounded iteration loop to test convergence.
- Returns an error if the conversion cannot converge.
Input and Result Behavior
Section titled “Input and Result Behavior”- Changing any parameter clears previous results (prevents stale output).
- Real-time validation flags non-numeric entries.
Copy Behavior
Section titled “Copy Behavior”- Results are copied with full floating-point precision.
Troubleshooting & Known Limitations
Section titled “Troubleshooting & Known Limitations”Weibull Mean/SD May Fail
Section titled “Weibull Mean/SD May Fail”Cause: Coefficient of variation is unrealistic.
Recommendation: Use the Shape & Median instead.
Log-Normal Constraints
Section titled “Log-Normal Constraints”- Mean and median must be positive.
- Zero or negative values will produce invalid results.