Release history Changes in each release of Analyse-it

MSA

• Changed default Linearity command to CLSI EP6-AEd2 options

Fit Model

• Added Cox proportional hazard regression model.
• Improved odd-ratios in logistic regression to support models including interaction terms.

Multivariate

• Added support to PCA monoplot to vary vector colors.

Fit Model

• Logistic/Probit fit tighter convergence criteria and added warning message when a large number of fitted values p=0 or p=1.

MSA

• Fix so variance function uses measuring interval for calculating fit

Added Recalculate All command to recalculate all analyses in the workbook.

MSA

• Added support for the new CLSI EP6-Ed2 linearity procedure. CLSI now defines the deviation from linearity as the difference between the mean of replicates and the linear fit, and therefore no longer requires fitting polynomial models. (Estimating the linearity of a measurement procedure (EP06-Ed2))

Minor bug fixes

Minor bug fixes

Fit model

• Probit/Logit regression inverse prediction confidence intervals

Diagnostic performance

• Noninferiority/Equivalence tests for comparing sensitivity/specificity
• Updated help topics for comparing sensitivity/specificity

Change to free trial

Fit model

• Probit regression (including features to perform CLSI EP17-A2 using Probit regression) (Fitting a simple probit regression, Estimating the detection limit using a probit fit)
• Transform X and Y variable in simple regression

Distribution

• Transform variable

MSA

• Predict variance at a given U for precision profile variance function
• Improved handling of small negative numbers in fitting a precision profile variance function (thanks to Bill Sadler for his suggestions)
• Hide excluded outliers in Variability plot

Method Comparison

• Set seed in P&B bootstrap procedure
• Log X-axis scale on difference plots

Reference Interval

• Set seed in bootstrap procedures

Diagnostic test

• Multiple prior probabilities for computation of predictive values.

MSA

• Changed Detection Limit non-parametric quantile computation method to align with CLSI EP17

Added EP06-A tutorial

Linearity

• Weighted regression fit for scenarios where the precision is not constant across the measuring interval
• Adjust the measuring interval to find the linear range

Precision

• Removed rounding of Satterthwaite DF for better compatibility with other software and EP05-A3 examples
• Fixed computation of confidence interval on partial sum of variance components

Improved quick-start guide

Included example datasets and analyses

Fixes for Office 2016 monitor DPI awareness

Temporarily fixes a problem where the Analyse-it user interface does not appear in the task pane on Excel 2016. A permanent fix will be available as soon as possible.

Improvements to the Ultimate & Method Validation editions:

Added EP05-A3 tutorial

MSA

• Added Generalized ESD outlier identification (Grubb’s test for more than one outlier) to MSA precision analysis.
• Improved Variability plot to show possible and excluded outliers.

Added EP15-A3 tutorial

Method comparison

• Various improvements to prediction interval in Deming regression

Fixed start-up issue experienced by some customers

Precision

• Added 4-parameter flexible variance function for precision profiles with a turning point near the detection limit

Round-up of recent minor fixes

Method comparison

• Fix for Mean vs 1st replicate computing lambda for Deming regression with unequal replicates per item.

Initial test release for .NET 4.0

Analyse-it Method Validation edition, version 4.65

Method comparison

• Added vary color / symbol by to plots allowing identification of stratification
• Added mean of X vs 1st Y replicate for compatibility with version 2
• Added Sy.x to Deming regression for compatibility with version 2

MSA

• Changed colors on variability plot to be more distinctive

Fit Model

• Changed vary color / symbol by to apply to residual and other diagnostic plots in addition to the scatter plot

Fix to linearity scatter plot

Release candidate of the Analyse-it Quality Control and Improvement edition, see http://analyse-it.com/blog/2015/10/announcing-the-analyse-it-quality-control-improvement-edition

• Improved: Ribbon icons

Reference Interval & Diagnostic performance

• Combined onto a single menu

MSA

• Added tool-tips
• Fixed assigned values sometimes could not be assigned to levels on MSA

Method Comparison

• Added non-parametric Bland-Altman limits of agreement tutorial

Distribution

• Added option for 1st and 99th percentile

Support for dates in categorical data

Method Comparison

• Added Bland-Altman limits of agreement based on ratio of measurements, linear regression, and non-parametric percentile method.
• Added in-depth tutorial on how to establish limits of agreement in various different situations.

Method Comparison

• Added mountain plot

Minor change to floating-user licences for use on virtual desktops.

Minor bug fixes for Excel 2013.

Bug fix for co-existence with Analyse-it v2.xx

Bug fix for excessive CPU use before concurrent-user licence is acquired

Major new release of the Analyse-it Method Validation edition.

Method Comparison

• Method comparison functions: scatters plots, difference plots, and fits, are now in one place so it’s easy to make decisions or choose fits after looking at the data.
• Ordinary linear, Weighted linear, Ordinary Deming, Weighted Deming and Passing & Bablok regressions.
• Constant difference and relative difference plots with mean difference, allowable difference, and limits of agreement.
• Bootstrap confidence intervals for parameters and bias in Passing & Bablok regression.
• Exact p-values on CUSUM linearity test.
• Predict mean difference at decision levels.
• Equality and equivalence hypothesis tests at decision levels.
• Kappa and Weighted Kappa qualitative method comparison.

Precision

• Graybill-Wang / Ting-Burdick-Graybill-Jeyaratnam-Lu modified large sample confidence interval estimators for variance components – produces confidence closer to the stated level than Satterthwaite confidence intervals, which can be too liberal.
• Support for unbalanced 1- and 2-factor nested random ANOVA models, for looking at precision between days, runs, laboratories, etc.

Linearity

• Hsieh-Liu confidence interval estimator for degree of nonlinearity.
• Equality and equivalence hypothesis tests for nonlinearity.

Reference interval

• Harrell-Davis, Bootstrap, and Robust Bi-weight quantiles.
• Box-Cox and other transformations.

Diagnostic performance / ROC

• Compare up to 10 independent or correlated ROC curves.
• Equality, equivalence and non-inferiority hypothesis tests.
• Estimate false positive fraction (FPF) at fixed sensitivity, sensitivity at fixed FPF, and sensitivity/FPF at fixed cut-off.
• Find optimal decision threshold based on costs.

Binary diagnostic tests

• Compare sensitivity and specificity of 2 independent or correlated tests.
• Diagnostic odds-ratio and Youden’s index.
• Improved confidence interval estimators for likelihood ratio and predictive values.

Changed analysis commands to always show pop-up menu for access to starting points.

Minor change to avoid buffer overrun warning with Sophos anti-virus.

Major new release of the Analyse-it Standard edition.

• Supports Excel 2007, 2010, & 2013, 32-bit and 64-bit versions
• Improved user interface for faster interactive analysis

Distribution

• Histogram, polygon, cumulative histogram
• Univariate box-whisker, dot-, and mean-plots
• Distribution plots normal Q-Q, CDF plot
• Shapiro Wilk, Kolmogorov Smirnov, Anderson Darling normality tests
• Z-test, t-test, Wilcoxon, Sign
• Binomial test, Pearson X², Likelihood ratio G²
• Frequency plot, whole-to-part plots

Compare groups

• Univariate box-whisker, dot-, and mean-plots
• Z-test, t-test, Welch t-test
• Wilcoxon Mann-Whitney test, Kruskal-Wallis test
• ANOVA, Welch ANOVA, Multiple comparisons
• Fisher F-test, Bartlett, Levene, Brown-Forsythe test for homogeneity
• Fisher exact, Pearson X², Likelihood ratio G²
• Mosaic plot

Compare pairs

• Univariate box-whisker, dot-, and mean-plots
• Difference plot, histogram of differences
• Z-test, t-test
• Wilcoxon test, Sign test
• Frequency plot

Correlation

• Scatter plot with density ellipse
• Scatter plot matrix with density ellipse and histograms
• Pearson, Spearman and Kendal correlation coefficient

Fit Model

• Linear and logistic regression
• Simple models: linear, logarithmic, exponential, power, polynomial
• Advanced models with simple, crossed, polynomial, factorial terms, with categorical variables coded as dummy variables
• Leverage, residual, outlier & influence diagnostic plots
• Standardized & studentized residuals, Cook's D
• Test effect of each term, test effect of model
• Predict Y for X

Minor fix for blank legend names on latest Excel 2007/2010 service packs

Minor update to fix an issue on start-up with Excel 2010 with Service Pack 1 and the Office File Validation add-in.

Fixed CUSUM linearity test on Passing-Bablok

Fixed significance test on lack of fit for non-linear model on Linearity

Compatible with Excel 2010 (32-bit only)

• Added output of fit equation to linear & polynomial regression output

Fixed occasional random crashes for users of Excel 2007.

• Fixed percent style re-definition warning that sometimes appeared on report output.
• Fixed box-plots drawing where connecting lines sometimes appeared.
• Change to rounding of degrees of freedom in CLSI precision.
• Added option to suffix name of diagnostic test with /H or /L to force ROC algorithm to treat higher/lower values as abnormal.
• Added UpdateReport function to allow updating an Analyse-it report from VBA.

• Fixed issue with significant figures formatting on the regression equation (on the scatter plot) for very large coefficients.
• Fixed Kruskal-Wallis contrasts p-value to a maximum of 1 (Bonferroni method can, in rare cases,calculate p-values larger than 1).
• Fixed cosmetic issue, due to a bug in Excel 2007 charts, where the horizontal X axis appeared sloped.
• Update network license no longer requires re-activation

• Added check to ensure all observations for Weighted Deming regression are greater-than zero.
• Changed ROC AUC confidence interval to be bounded between 0..1. No changes made to the AUC confidence interval calculation.
• Fixed label on regression output that shows the confidence level %, which always showed as 95% even if a different level was requested. The confidence intervals calculated are correct, and no changes made to the calculations.

• Fixed code-signing certificate expiry issue.
• Statistics in chart legends are now formatted to same precision as the source variables.
• Fixed display issue, where an advisory label indicating when methods compared using Passing-Bablok have a non-linear relationship (p-value > 0.10) was not shown.

• Fixed labelling of > and >= cut-off operators were reversed, performing >= and > respectively, on the Qualitative Test Performance in Method Evaluation edition.

• Fixed regression residual histogram scaling issue on Microsoft Excel 97-2003.

• Fixed chart axis label positioning issue that appeared on some non-English language versions of Microsoft Excel.

• Kolmogorov-Smirnov now uses the method of Dallal & Wilkinson (The American Statistician, 1986, Vol 40, No 4, pg 294-296) for computation of p-values. The new method is more accurate for small sample sizes.
• Kolmogrov-Smirnov and Anderson-Darling now report the D and A2 statistic directly rather than the sample stabilized A* and D*.
• Added 1st and 3rd Quartile to comparative descriptives output.
• Added TP / TN / FP / FN to ROC curve output.

Added Getting Started tutorial.

True Excel 2007 interface, with quick access to tests and tools.

• Simpler dataset setup with predefined experiment templates.
• Improved histograms, box-plots, mean-, dot- and scatter plots.
• Major improvements to all the statistical tests, especially the Method Evaluation edition.
• New easy-to-read online help, with a new Getting Started tutorials.

Existing users are encouraged to update to v2.

Made Kappa test available to the General statistics module. Added an option to omit confidence intervals on comparative box-plots, useful when the sample size is small and box plots appear 'inside-out'.

Changed Passing-Bablok method conversion (regression) algorithm to handle ties in Y.

Modified plotting of ROC curves slightly so the end of the curve is always joined to (1,1)

Office XP compatible

Improved license system and installer. Existing users upgrading to version 1.60 or later must request a license key (see below).

Added linearity protocol. No documentation is included for this procedure as yet, but an example called "Linearity.xls" is included in the "Analyse-it\System\Examples" folder. If you have any questions please contact us.

Changed linear regression prediction intervals to use t-distribution instead of F-distribution. Warning: the new t-distribution implementation yields narrower prediction intervals compared with the F-distribution.

Added "no intercept in model" to linear regression allowing fitted regression line to be forced through 0 (a.k.a. zero forcing).

Corrected problem where confidence level was not shown on linear regression. Confidence intervals were correct, but confidence level was always shown as 95% instead of chosen level.

Fixed bug in 2-run precision analysis when comparing within-run observed precision against precision claim.

Modified Deming regression residual plot to use standardized residuals instead of absolute residuals to allow easier spotting of outliers (use standard +/- 4 SD rule).

Added 'Linear method comparison' procedure -- basically linear regression modified to be better suited for method comparison studies (p-value tests slope against a hypothesis of 1).

Fixed bug in non-parametric reference intervals, where upper confidence level was reported even though too few samples (less than 119 samples).

Improved ROC analysis, with tables showing sensitivity/ specificity/predictive-values for all possible decision-thresholds of each diagnostic test.

Improved linear regression algorithm to produce more accurate results when dependant variables are highly co-linear (now results are more accurate than Minitab!)

Modified precision plot, and fixed bug in the %difference plot of bias plots.

Modified ROC analysis to automatically decide whether higher or lower diagnostic results indicate normal or abnormal.

Added residual plots to Deming and Passing-Bablok method comparison procedures.