| Method validator - user manual |
1- Description of the programme
Method validator is a statistical programme designed for comparing analytical techniques in clinical laboratories. Its aim is to provide a simple tool which can be used intuitively without any training. It makes use of conventional statistical techniques :
The graphical tools have been studied in depth to simplify the plotting of data. Import/export functions are provided to allow exchange of data with a spreadsheet or copy/pasting of the graphs in a word processor.
A random number generator allows various experimental situations to be simulated by changing the analytical CV of each method and the distribution of data.
The difference plot proposed by Bland et Altman [2], is the comparison technique recommended by the "Annals of Clinical Biochemistry" [3]. Log transformation of data is often required in clinical laboratories when the analytical error increases proportionally with the measured value (constant CV).
Linear regression is based on an unsuitable theoretical model in which the independent variable is assumed to be measured with no error. It is better suited to calibration operations than to method comparisons. The Linnet weighting [5] corrects heteroscedasticity error for a constant CV.
The weighted or unweighted Deming regression takes into account the variance of the two methods. It introduces an additional parameter: the ratio of the variances of the two methods. This ratio is usually known, since the first stage of an analytical method evaluation is the determination of its CV. The confidence intervals are calculated using the "jackknife" algorithm [4].
The Passing and Bablok methods, in common with all non-parametric methods, are less sensitive to outliers. Two techniques are provided : the first [8,9] is a test of the agreement of two analytical methods, the second [1] is a technique for determination of the regression parameters.
Note : the calculation time and system resources required for the Passing-Bablock method increase very rapidly with the number of experimental points. A delay of a few seconds before the results are displayed is normal when there are several hundred points.
The programme displays two main windows :
The "Data entry" window cannot be re-sized. When visible, it always stays on top of other windows. It can be hidden or re-displayed using the <F12> function key. The "Comparison plots" window can be re-sized as required including full-screen display.
The active window is the one which receives the keyboard commands. It can be identified by the upper blue title bar, whilst the inactive window is grey. To switch between windows, simply click anywhere in the window or press the <Ctrl + Tab> keys together.
When the programme is first started after installation, a random data set is automatically generated in order to facilitate initial use of the programme. Analytical data can be subsequently entered via the keyboard, loaded from a file or pasted from the clipboard.
Method names
Two fields are provided for entry of the names of the two methods to be compared (maximum 15 characters). These names are printed on the reports. The default name of the reference method is "Ref. method" and that of the method to be evaluated is "Field method".
Statistics of entered data
The lower table gives a summary of the data entered for the two analytical methods (mean, standard deviation, minimum and maximum values, number of data points and correlation coefficient). These statistics are updated after the <Apply data> button is clicked.
Validation of entered data
The data entered via the keyboard, imported (as *.txt file) or pasted from the clipboard are not immediately incorporated. The <Apply data> button must be pressed first. All of the data in the table are then applied. If one of them is incorrect, a warning is displayed. If all of them are correct, the "comparison plots" window is updated.
Location of a point from a line in the table
It is often useful to locate a point on the graph from its co-ordinates. This can be done by clicking on the relevant line number in the data entry table. This number and the associated point are displayed in red. The reverse operation is possible : when one of the points in the graph is clicked, the corresponding line in the table is shown in red.
Files
The data are saved in specific Method Validator files bearing the extension ".mvd". These files archive the co-ordinates of the points, and the settings for printing the graphs : colours, method names, title, character sizes, shape and size of the points.
The name of the current file is displayed at the bottom of the "Data entry" window. The default file names are MethValXXX.mvd where XXX is an incremental number starting at 001 and depending on the numbers already assigned in the working directory.
5- Data entry via the keyboard
Navigation in the data entry table
New data pairs
To add a new data pair, simply enter the two values for X and Y on a single line in the table. The added data are not incorporated until the <Apply data> button is used. The addition can be made to any empty line in the table. Any empty lines will be re-ordered when the the <Apply data> button is used to remove any spaces. Either decimal points or commas can be used.
Deletions and corrections
6- Menus in the "Data entry" window
File menu
The <File> menu in the data entry window enables data to be loaded or saved. Two types of files are available :
Clipboard menu
Data menu
Print menu
? menu
When the "Comparison plots" window is re-sized, the size of the data points and the font size are automatically re-calculated depending on the smallest dimension of the frame of the plot. If the "Data entry" window interferes, it can be hidden (then re-displayed) with the <View|Show-hide data entry> menu or with the F12 function key.
Statistical method
The "Comparison plots" window provides 4 statistical methods for data analysis. Select the method required by clicking on the relevant tab. The upper frame shows the results of the data analysis. Check boxes allow various calculation options to be selected (logarithm, weighting, agreement/estimation test), depending on the statistical method.
Location of the co-ordinates of a point
When a data point seems to be an outlier, it may be useful to check its co-ordinates. Click on the suspect point. The point and the number of the line in the data table are then displayed in red. It is then easy to delete the point or correct the co-ordinates.
8- Menus in the "Comparison plots" window
View menu
Print menu
? menu
This window allows simple setup of the page and graphs. These can then be printed immediately or copied to the clipboard for transfer to a word processor.
Position of the graph on the sheet
The most accurate method requires setting of the margins (cm) in the <left and bottom margins> fields. The position of the graph on the sheet can be also easily adjusted using the mouse :
Pre-defined settings buttons :
Size of the graph on the sheet
The most accurate method is to set the dimensions (cm) in the <width> and <height> fields. The size of the graph can be also easily adjusted using the mouse. Move the arrow of the cursor to the red square at the top right of the graph containing an arrow with two points, hold down the left button of the mouse and drag the mouse to re-size the graph.
Pre-defined setting buttons :
Printing of the title
<Graph title>: Entry field for the graph title. This is only updated after leaving the field. The title is printed in bold on the first line of the caption.
The caption is printed at the bottom centre of the graph. The size of the characters can be adjusted with the <Caption font size> trackbar. After each modification, the graph is re-sized to occupy the available area. Three choices can be made by checking the <Caption> boxes :
Please note that captions which are wider than the frame of the graph are truncated to the width of the graph for the copy/paste function. This can be easily checked beforehand using the zoom button. The width of the captions is retained if the paper size allows when the Method Validator programme is used to print the graph directly.
Other settings
Paper format
The < Change> button in the "Paper size" frame allows :
The printable area displayed in the <Paper dimensions> frame is slightly less than the paper size due to the unprintable edges of the paper. The maximum dimensions are 32 cm (limit imposed by the Windows "extended metafile" format). The US legal format cannot be used, therefore, due to its height (14 inches = 35.6 cm).
Printing or copying to the clipboard
The colours in each of the "Comparison plots" and "Preview" windows can be changed separately. The dialogue is achieved with the same set of buttons in both cases :
The <Default> button restores the default
colours.
The <Undo> button cancels all of the changes made since the
dialogue was started.
When the programme is closed, a configuration file (MethVal.cfr) is copied to the disc in the current directory. It contains the choices made on completion of the session which will be restored to the following :
The random points generator creates random independent (X,Y) pairs which are normally distributed. Its aim is to instruct by facilitating proficient use of the proposed statistical techniques. The calculation of each pair is achieved in four stages :
The number of random points can be between 1and 1000.