The MLPowSim software is now available as version 1.1. We must stress here that MLPowSim is free software and comes with no WARRANTY whatsoever. Below you will find the software and manual along with two accompanying datasets. The manual has been written by William Browne, Mousa Golalizadeh and Richard Parker with the software written by William Browne and Mousa Golalizadeh.
MLPowSim was funded by the ESRC Grant: Sample size, Identifiability and MCMC Efficiency in complex random effect models. Further details can be found at Sample size and MLPowSim.
- Request the MLPowSim executable file
- MLPowSim manual (PDF, 2mb)
- A few talks have been given on the topic of sample size calculations in multilevel models.
Major changes and bug fixes since the initial release
- Turn off exclusion column at the end of the simulation loop to avoid wrong column length error when next set of data are simulated
- Fix to work with recent version of MLwiN where overwriting protected columns is prevented
- Save worksheet space by saving just the standard errors, instead of the full covariance matrix in simulations
- Avoid reading from the keyboard directly when requesting the user "Press any key to continue" so that the command can be scripted without user intervention
- Increase total columns in the worksheet, and dynamically assign these to outputs in order to increase the maximum number of power parameters beyond the original 10
- Correct R code generated for three level unbalanced cross-classified models for unbalanced scenarios 1 and 2
- Update R code to account for changes in lme4 package options
- Add quotes around the link function in the generated lme4 R code
- Attempt to improve consistency of the question formatting between the different model types
- Attempt to improve indentation and formatting consistency in the generated scripts
- Add checks of the validity of input values as the user enters them
- Switch to using EXCL command in generated MLwiN code to create unbalanced cases
- Replace response with a column of zeros before simulating in MLwiN code so that previously missing rows in the response do not exclude them from the next simulation
- Remove unnecessary commands from the generated MLwiN code
- Differentiate between integer and decimal inputs and ensure the user has provided the expected type
- Within the generated MLwiN script clear the model before setting up the next simulation inside the innermost loop to ensure left-over information doesn't cause problems
- For MLwiN cross-classified models, turn on cross-classification flag after the initial model set-up and off after running the main loop
- Avoid crash when generating MLwiN code for 2 and 3 level unbalanced nested models by correcting a matrix allocation size
- Check values of matrices are valid as the user enters them
- Use consistent indentation in generated MLwiN scripts
- Handle 0/1 case where the interval is entirely within the negative range
- Remove unnecessary code from the generated R scripts
- Update the user manual to reflect changes to MLPowSim, MLwiN, R and lme4
Bugs in MLwiN version 2.10 that apply to code generated by MLPowSim
- A bug in the multivariate normal random number generator causes a crash when running code generated by MLPowSim containing the MRAN command. Fixed in version 2.11.
- The 1 level Binomial and Poisson models throw up an error message in MLwiN. This can be cured by replacing the file PRE in the discrete sub-directory of the MLwiN install with the following version PRE. Fixed in version 2.11
Other Multilevel Sample Size software
- The PinT software (Tom Snijders, Roel Bosker, and Henk Guldemond) used for comparison in the manual.
- The ML-DEs software package (Cools, Van den Noortgate and Onghena) that has been developed independently from MLPowSim but which also uses MLwiN and simulation to calculate power calculations for multilevel designs is available. We hope to compare MLPowSim with ML-DEs in further work.
- The OD (Optimal Design) software package (Steve Raudenbush and colleagues) also looks at multilevel power calculations and in particular cluster randomized designs.
Note: some of the documents on this page are in PDF format. In order to view a PDF you will need Adobe Acrobat Reader