PASS 2020 Update

Update Release Notes

This page lists the changes that have been made to PASS 2020 since it was released.

Version: 20.0.10

Released: January 23, 2023

1. In the Confidence Intervals for One-Way Repeated Measures Contrasts procedure, there is a Multivariate option and a Univariate option. A bug was causing an error whenever the Univariate option was selected. Now, both options can be run with no error.

2. In the Mann-Whitney U or Wilcoxon Rank-Sum Tests for Non-Inferiority and Mann-Whitney U or Wilcoxon Rank-Sum Tests for Superiority by a Margin procedures, the sample sizes were not reported correctly in the summary statements when solving for effect size. Now, the sample sizes are consistent with those shown in the table.

3. In the Non-Inferiority Tests for the Difference of Two Means in a Higher-Order Cross-Over Design procedure, if "Specify sigma as sigmaw or sigmab and row" was set to "sigmab and row", the sigmab did not read in correctly and the procedure produced incorrect results. This was fixed to read in the sigmab value, and the results are now correct.

4. In the six group-sequential survival (simulation) procedures, the accrual parameter, if different from 0, was not implemented correctly. If the accrual parameter was other than 0 (or the "Percent of T0 Until 50% are Accrued" was other than 50), the starting times for the simulations were generated incorrectly. This was fixed so that the start times are distributed according to the truncated Exponential distribution, as indicated in the option info.

5. Corrected an error in power calculation for the Equivalence Tests for Two Means in a Cluster-Randomized Design procedure. When Test Statistic was set to “T-Test Based on Number of Clusters” (not the default) and K1 was not equal to K2, the power was calculated using the wrong degrees of freedom. This resulted in the power being slightly off the correct value.

6. This is the final update to PASS 2020.

Version: 20.0.9

Released: August 9, 2022

1. Corrected the formula for the SD of the Uniform Distribution in the documentation for the Data Simulator procedure. The procedure calculations were not affected.

2. Fixed the Accrual Time (Integers Only) input option in the Two-Group Survival Comparison Tests (Simulation) and Group-Sequential Logrank Tests (Simulation) (Legacy) procedures so that decimal numbers are not included as possible options in the dropdown. The procedures were also updated so that decimal number are correctly rounded to whole numbers before calculation.

3. Corrected an error in the Non-Zero Null Tests for Simple Linear Regression using ρ² procedure. For some combinations of ρ0² and ρ1², solving for N (Sample Size) would return no results.

4. Corrected an issue in the Confidence Intervals for Kappa procedure. When Interval Type was set to something other than Two-Sided, the software was not reading the Distance from κ to Limit option correctly. Instead it was reading and using the hidden Width of Confidence Interval option value in its place.

5. In the Confidence Intervals for an Exponential Lifetime Percentile procedure, the Width of Confidence Interval was limited to values between 0 and 2. It was fixed to correctly allow for all values greater than 0.

6. Corrected an error in the following procedures: Tests for Two Means in a Repeated Measures Design, Tests for Two Proportions, Fisher's Exact Test for Two Proportions, Non-Inferiority (Superiority by a Margin, Equivalence, Non-Zero Null) Tests for the Difference (Ratio, Odds Ratio) Between Two Proportions, Tests for Two Independent Sensitivities, Non-Inferiority (Superiority by a Margin) Tests for Vaccine Efficacy using the Ratio of Two Proportions. When solving for sample size with Group Allocation set to “Enter R” or “Enter percentage in group 1”, the reported sample sizes might have been slightly off (usually by only 1 subject) due to incorrect rounding of the sample sizes.

Version: 20.0.8

Released: February 15, 2022

1. Modified the allowable ranges for Cluster Size parameters M and m in the three Stepped-Wedge Cluster-Randomized Design procedures. The restriction that both parameters be ≥ 2 was found to be unnecessary. Now both parameters have allowed range of > 0.

2. An error in the “Mann-Whitney or Wilcoxon Rank-Sum Tests (Simulation)” and the “Tests for Two Means (Simulation)” routines has been corrected. The error caused the mean differences to be reported incorrectly in cases where the distribution’s first parameter was not the mean. The power calculations were correct.

Version: 20.0.7

Released: January 20, 2022

1. Corrected an error that caused the text in some single-value dropdown inputs on the Procedure Window not to be read correctly after selecting an item from the dropdown list.

2. Corrected an error in the Non-Inferiority (and Equivalence) Tests for the Difference (and Ratio) of Two Correlated Proportions procedures. When one of the cell proportions was zero, the software would incorrectly indicate an undefined table in some cases.

3. Corrected an issue with graphics dropdown buttons when using Windows 11. Dropdown buttons on plot format windows would display two images (one slightly offset from the other).

4. With this update, PASS 2020 is considered to be fully compatible with Windows 11.

Version: 20.0.6

Released: August 30, 2021

1. Corrected an error in all GEE Tests procedures. The correlation matrix was not computed correctly when using Damped Exponential as the Pattern of ρ’s Across Time.

2. Corrected an error that occurred when parsing user-entered TO BY lists (e.g., "-2 to 2 by 0.1"). In some cases, rounding errors were causing unexpected results.

3. Corrected an error in the Superiority by a Margin Tests for Two Means in a Cluster-Randomized Design procedure when solving for M1. The search now converges.

4. Corrected an error in the Tests for Two Proportions in a Stepped-Wedge Cluster-Randomized Design procedure when solving for Effect Size with Effect Size Input Type set to Differences, Ratios, or Odds Ratios. The search now gives the correct result.

5. Corrected an error in the Tests for Two Means in a Stepped-Wedge Cluster-Randomized Design procedure when solving for Effect Size with Effect Size Input Type set to Differences. The search now gives the correct result.

6. Corrected an error in the Tests for Two Poisson Rates in a Stepped-Wedge Cluster-Randomized Design procedure when solving for Effect Size with Effect Size Input Type set to Differences or Rate Ratios. The search now gives the correct result.

Version: 20.0.5

Released: February 19, 2021

1. Minor documentation updates.

2. Corrected an error in the Logrank Tests procedure. When using Mortality input, power and sample size were not computed correctly when entering M1 (Mortality - Control) and MR (Mortality Ratio = M2/M1). Power and sample size were correct when entering M1 (Mortality - Control) and M2 (Mortality - Treatment).

3. Corrected “Unhandled exception…” errors that occurred when opening PASS procedures in Windows 10 when the Microsoft Region language option Beta: Use Unicode UTF-8 for worldwide language support is active. (To determine if the option is active on your computer, see the steps below.)

Having the Beta: Use Unicode UTF-8 for worldwide language support option active may lead to other errors. We recommend that the option be disabled when using PASS.

To disable this option in Windows 10, perform the following steps with administrator rights on your computer:

a. Close all open documents and programs on your computer.
b. Press the Windows key.
c. Type “Settings.”
d. Select Time & Language.
e. Click Language.
f. Under Related settings, click Administrative language settings.
g. On the Region window, click the Administrative tab.
h. Click Change system locale. (Here you will need administrator rights.)
i. Clear the Beta: Use Unicode UTF-8 for worldwide language support checkbox.
j. Click OK.
k. Restart your computer.

Version: 20.0.4

Released: January 28, 2021

1. Corrected an error in the two Mendelian Randomization procedures that caused the reported value for N when searching for sample size to be off by 1 in some cases.

2. Corrected an error in the Tests for Two Ordered Categorical Variables procedure. The software was not displaying the Group 2 Proportion values correctly (it was displaying all zeros for P2. The error did not affect calculations.

3. Corrected fatal errors that were occurring when editing lines, colors, and symbols in Group-Sequential Boundary Plots.

4. General startup stability improvements.

Version: 20.0.3

Released: August 27, 2020

1. Mediation procedures have been corrected to allow for negative value input in all appropriate cases. Before the corrections were made, the sample size search would not converge.

2. Corrected a “value out of range” error that was occurring in the Confidence Intervals for One Proportion in a Stratified Cluster-Randomized Design procedure in some cases when the computer language setting uses comma as the decimal.

3. Corrected an error in the display of input headings for the One-Sample Logrank Tests procedure. The "Alpha" heading was being displayed in error on the panel when the Solve For was set to "Effect Size".

4. Corrected output typos in the two-sided null hypothesis for the Two-Sample T-Tests Assuming Equal Variance procedure, Two-Sample T-Tests Allowing Unequal Variance procedure, and the Mann-Whitney U or Wilcoxon Rank-Sum Tests. The null hypothesis for the two-sided test was incorrectly stated as H0: δ ≠ 0 instead of H0: δ = 0. This error did not affect calculations.

Version: 20.0.2

Released: May 22, 2020

1. Updated the Tests for Two Ordered Categorical Variables procedure to include Odds Ratio input (in addition to Log(Odds Ratio) input) and additional methods for specifying the control group category proportions.

2. Added examples to demonstrate how the Tests for Two Ordered Categorical Variables procedure can be used in planning a clinical trial to assess the effectiveness of a treatment in combatting COVID-19 where the study outcome is based on a multi-category ordinal scale of illness severity.

3. Added an example to demonstrate how the Logrank Tests (Freedman) procedure can be used in planning a clinical trial to assess the effectiveness of a treatment in combatting COVID-19 where the study outcome in each group is the number of days it takes patients to advance two categories on a multi-point ordinal scale of disease status.

4. Corrected an error in the Equivalence Tests for the Ratio of Two Means (Normal Data) procedure. The values of N2 and N were displayed wrong in some cases. Incidentally, N2 is not used in the power calculation, so that value is correct.

5. Corrected an error in the Group-Sequential simulation procedures that required the user to repeatedly press “Abort” in order to Abort a procedure run.

6. Corrected an error in the Cox Regression procedure when solving for power with B = 0. PASS was returning Beta for Power instead of returning Alpha for Power.

7. Corrected an error that caused plots with Extra Symbols added outside the bounds of the plot lines to become tiny. Extra symbols outside of the plot boundaries are not displayed.

8. The validation example (based on Lakatos, 1988) for the Two-Group Survival Comparison Tests (Simulation) procedure and the Logrank Tests procedure had the noncompliance and drop-in rates reversed. This was fixed in the documentation and in the validation example templates. The fix did not result in a change to the estimated sample size.

9. Corrected minor typos in the power formula in the documentation of the Equivalence Tests for the Difference Between Two Correlated Proportions procedure.

Version: 20.0.1

Released: February 10, 2020

1. Initial Release of PASS 2020.