Software changes

PanelCheck 1.4.1: Changes from previous version PanelCheck v. 1.4.1:

(March 08, 2015)

Bug-fix 1: in the p-plot (see tab named "F & p plots") the colouring of lines for p-value did not correspond correctly with the colours shown in the legend (also in the same plot).

Bug-fix 2: in the numerical results of ANOVA in the tab named "Overall" a few lines have been removed (those named "F prod", "F ass*prod interaction" and "F prod / F ass*prod interaction"). All numbers shown are correct (we have checked them really many times since ANOVA was implemented), only colouring of one line was wrong. We removed those lines, since they really were redundant.

PanelCheck 1.4.0: Changes from previous version PanelCheck v. 1.3.2:

(September 03, 2010)

Please find detailed information on changes in file 'PanelCheck 1.4.0 - new features and changes.pdf'. Download it here.

PanelCheck 1.3.2: Changes from previous version PanelCheck v. 1.3.1:

(July 21, 2008)

New features for enhanced useability:

    • 3-way ANOVA is now 'complete' by supplementing it with analysis and plots for replicate effect, assessor*replicate interaction and product*replicate interaction.

    • Precise p-values in 2-way ANOVA (1 rep), 2-way ANOVA and 3-way ANOVA instead of categories as p<0.001, etc.

    • Possible to run 2-way ANOVA (1 rep), 2-way ANOVA and 3-way ANOVA for one attribute at a time

Bug-fixes:

We always test all our methods thoroughly and try to catch and remove bugs in computations if they should occur. However, sometimes a bug may unfortunately go undetected. This has been fixed from the previous version (1.3.0):

    • In some few cases LSD bars in 2-way ANOVA (1 rep), 2-way ANOVA and 3-way ANOVA had wrong colour of significance.

PanelCheck v. 1.3.0:

(june 09, 2008)

New methods and plots:

    • ANOVA: after many requests 2-way ANOVA for one or more replicates has been implemented for your convenience. You will find the methods under the 'Overall' tab. What was called 'Mixed model ANOVA' in the previous release is now named '3-way ANOVA'. Many thanks to our danish collaboration partners and friends from DTU for contributing the code.

    • Manhattan plots for screening purposes and quick overview over systematic variance for assessors and attributes. More information on Manhattan plots can be found here.

    • Bi-plot for PCA on consensus average (original and standardised) and STATIS compromise

    • Spiderweb (radar) plot for consensus average (original and standardised) as well as STATIS compromise

New features for enhanced useability:

    • Coloured frames for plots with focus on attributes: coloured frames indicate whether the differences between the tested products is significant for the attribute at hand. The colous are based on product significance from 2-way ANOVA.

    • Missing data:

    1. Summary about missing data (how much data is missing for each assessor, etc)

    2. The missing value is replaced with either the column mean or row mean of the remaining scores. The mean with the smaller standard deviation is inserted. This procedure is carried out assessor wise.

    • File loading:

    1. Recent file history for faster loading of recently used data sets

    2. The data set used in the last session is reloaded automatically when starting PanelCheck next time

    • 2-way and 3-way ANOVA: LSD bars now can be dragged up and down, such that checking significance of product difference is done easier

    • Interactivity: double-click on points and bars in the following plots and a new relevant plot appear without having to switch between the plot tabs. These interactive plots are: all plots under the 'Overall' tab, common scores and correlation loadings plots under 'Tucker-1' tab, all PCA score plots under 'Consensus' tab and correlation plots.

    • Reduced plotting and computation times for overview plots

Bug-fixes:

We always test all our methods thoroughly and try to catch and remove bugs in computations if they should occur. However, sometimes a bug may unfortunately go undetected. Here are two bugs in the previous version (1.2.1) that we have fixed now. We apologise for any inconvenience.

    • With PanelCheck 1.2.1 in some cases (usually with scales ranging from 1 to 15) eggshell plots may return erronous results / product rankings. If you used scores with only one digit (i.e. from 1 to 9) this error does not occur. The error is not very likely to occour if you use for example a scale from 0 to 100.

    • Significance bars in LSD plots did not change colour according to significance level. They were always red and thus sometimes indicated that product signficance was at p < 0.001.

PanelCheck v. 1.2.1:

(june 11, 2007)

New methods and plots:

    • Mixed model ANOVA from our danish research fellows Per Brockhoff and Niels Sommer at DTU for overall evaluation of the panel. These 8 new plots are implemented under the new 'Overall'-tab in the graphical user interface (GUI).

    • STATIS-consensus under the 'Consensus'-tab, representing a better panel consensus with weighted assessor contribution. Implemented with the kind help of Dr. Mohamed Hanafi from ENITIAA, France.

    • Profile plot under the 'Univariate'-tab: plots the mean values of samples (attribute-wise) for the panel and for each assessor. Samples are sorted by intensity from left to right. This plot is especially useful for training sessions.

    • Mean & STD plots: Plots the mean and STD over samples for each assessor/attribute combination. Plots either sorted by assessor or attribute. These plots visualise how your assessors use the scale for an attribute.

New features for enhanced useability:

    • Smart data loading:

    1. Guided loading of sensory profiling data

    2. Columns in data file may be in random order, since user may select the specific columns during the import process.

    3. Select which columns shall be imported/not imported.

    4. Detection of unbalanced data.

    5. Trimming of unbalanced data to balanced data set

    6. Loading from text files works now for different delimiters: tab, colon and semicolon

    • Overview plots:

    1. Shows all plots of one type as subplots in a single frame for easier comparison and navigation through plots.

    2. Subplots in overview plot can be enlarged by clicking on them

    • New plot-handling features

    1. Zoom in/out

    2. Pan/move the axes

    3. Reset/restore plot after zoom or pan

    4. Copy-button for copying plot to clipboard (previously only possible by using ctrl-C)

    • Export

    1. Exporting multiple plots at the same time instead of using ctrl-c for one plot at the time

    2. Exporting plots directly to to PowerPoint-presentation

    • Restructured Help-documentation

PanelCheck v. 1.1.1:

(november 24, 2006)

Same functionality and features as PanelCheck 1.1.0. A few users reported problems about missing .dll-files when starting up PanelCheck 1.1.0. A new .exe-file for PanelCheck 1.1.1 was compiled including those missing .dll-files.

PanelCheck v. 1.1.0:

(november 21, 2006)

1. Added a new tab named 'Consensus', with sub-tabs 'original' and 'standardised' providing the following features:

Automatic calculations of consensus data matrices (averages) and PCA plots based on consensus data:

    • Calculate consensus matrix (sample averages over assessors and replicates) based on original scores

    • Calculate consensus matrix (sample averages over assessors and replicates) based on assessor-wise standardised scores: for each attribute within the data of one assessor: xstand = (x - xmean) / std(x)

    • PCA plots and results: scores, loadings, correlation loadings and explained variance plots for the averaged matrices

2. New buttons for easier navigation through plots:

The buttons “previous plot” and “next plot” allow you to browse through the plots without having to double-click on the tree-control every time when you wish to open a new plot.

3. New buttons for methods 'Tucker-1' and 'PCA' (consensus) :

With the new buttons you can now investigate also other PC’s other than PC1 and PC2. Just click on the arrows to select the PC’s you want to display.

4. Multiple selecting/deselecting of checkboxes:

If you need to select/deselect checkboxes for assessors, samples or attributes in a certain pattern (that is, every second, third, fourth, etc.) it is now possible to do so by choosing the selection pattern in the selection dropdown meny above the checkboxes.

NOTE: you'll have to click on the names instead of the checkboxes to select/deselect pattern-wise.

5. Enhanced line plots:

Different symbols for each replicate (replicate 1: circle, replicate 2: diamond, replicate 3: square, etc). Different colours for each assessor in the assessor-specific line plots.

6. Attribute names and sample names are now shown in line plots and eggshell plots, respectively.

7. PanelCheck is now registered in the window registry. That means that it can be opened by other programs (as for example by commercial software that collects the profiling data from your asssessors).

PanelCheck v. 1.0.3:

(july 30, 2006)

1. Fixed problem that gave erroneous F values for attributes where one assessor has zero variation.

2. Fixed bug that led to display of wrong raw data after pushing "Raw Data"-button in plot.

3. Colour enhacements in tree control, such that assessors and attributes are more distinguishable.

PanelCheck v. 1.0.2:

(june 18, 2006)

1. Fixed problems with data import from Excel.

2. PanelCheck remembers from which directory the last data set was loaded.

PanelCheck v. 1.0.1:

(may 25, 2006)

1. You can now drag and drop your data file onto the PanelCheck for easier data import

2. Improved robustness for import of data from plain text files and Excel files

3. Fixed bug where PanelCheck did not return ANOVA results (F plot, MSE plot and p*MSE plot) in one specific case