As always in experimental science, you need to analyze the data you collect to see what it's telling you about the world. ln the beginning stages, you usually need to be able to do some graphical analysis to get an overall view of any trends being revealed by your experiment. Qnce you get to the point of having a model you want to test, you then need to run some statistical tests to see how well your model fits the data. lt always is more convenient to learn one tool rather than two. To this end, there is the package called SciDAVis (Scientific Data Analysis and Visualization, http://scidavis.sourceforge.net). SciDAVis started life as a fork of QtiPlot. lt has moved quite a bit from the original codebase with the addition of several new features and changes to the underlying data structures. The functionality it provides is similar to commercial programs like Origin and SigmaPlot. lt also is similar to another open- source program called LabPlot. ln fact, beginning in 2008, these two projects started working together on a common back end while continuing with their own front ends and their own feature sets. ln this article, I take a quick look at some of the things you can do with SciDAVis for your own data analysis tasks. First, you need to install SciDAVis. Most distributions should have a package available. ln Debian-based distributions, you can install it with: sudo apt-get install sctdavts Binaries also are available for l\/lac OS X. If binaries aren't available, you can download the source tarball and build it specifically for your system. You need to have the Qt libraries installed, because the interface is built using them. When you first start SciDAVis, you get an empty project and an empty data table with two columns of numeric values (Figure 1). Selecting a column displays the details of that column in a window on the right-hand side. Selecting the description tab lets you change the name of the column, as well as
 | |
| Figure 1. Opening SciDAVis gives you an empty two-column data table. | | | |
add a comment describing what the column represents. The type tab lets you change what kind of data you can enter for this column from numeric to text, month names, day names or full dates. You also can set the format of the data type for each column. When you are first learning to use SciDAVis, you probably will just want some junk data to play With. You can do this easily by right- clicking the columns and selecting “Fill Selection with". Then, you can fill the column with either row numbers or random numbers. Qnce you have some data available, you can create new columns that are functions of the values stored in the other columns. To access this functionality, you need to select the formula tab in the right-hand side pane. ln order to explore some of the other features, let's set the first column to be the row numbers and the second column to be a set of random numbers (Figure 2). One of the first things you will Want to do is plot the data to see what it looks like. To do a basic plot, you simply can right-click on a column and select Plot. If you
 |
| Figure 2. You can use fill functions to create data in order to try things out. |
are just doing an initial look at the shape of the data, select Plot->Line or Plot-›Scatter (Figure 3). The x axis is simply the index values, and the y values are the data elements from the column. If you want to do more complicated plots, SciDAVis provides a full plotting wizard. To access the wizard, either press Ctrl-Alt-W or click View->Plot Wizard. This will pop up a new window where you can select which columns will be used for the x, y and z axes. You also have the option of selecting columns to represent the errors 24 / SEPTEMBER 2013 I WWW.LlNUXJOURNAL.COM for the x and y axes. You need to create a new curve where you can set the relevant columns. You can create multiple curves that will all be plotted on the same graph. You then get a new window with the plots generated. The point of SciDAVis is to make this type of work easy, so you can double-click on the various elements to edit the details of your graphs. Double-clicking on the title, or the axis labels, will pop up a window where you can change the content and the display. You also can change the details of the
 |
Figure 3. Scatter plots are only a couple mouse clicks away.
axes themselves. |
Double-clicking on an axis will pop up a new window where you can set the scale, the type of grid, the axis displays and a set of general options (Figure 4). Each of these sets of options is available under its own tab in the option window. Once the graph looks the way you want it, you will want a copy for your publications. To do this, you either can right-click on the graph and select Export or click on the menu item File->Export Graph. Then you can select your preferred file format for saving the image. Although graphs can be very useful when trying to get an intuitive grasp of the shape of your data, you do need to back up this intuition with hard numbers. The first thing to do, usually, is simply look at the column statistics. Right-clicking on the column, you would select Column Statistics. This creates a new table where you will get the number of rows in the column, along with other statistics like the standard deviation, variance, sum, minimum and maximum. You can see if there is any correlation between
 |
Figure 4. You can change all of the elements of your graphs.
two columns. |
You need to select two columns from your table, and then click on the menu item Analysis-›Correlate. This will pop up a new graph window showing a picture of the correlation. Two new columns will be added to your data table where you can find the lag and correlation values of this particular analysis. If the data you are looking at has some type of periodicity, you 26 / SEPTEMBER 2013 I WWW.LlNUXJOURNAL.COM can calculate an FFT of it to see the spread of frequencies within your data. By selecting a column and clicking on the menu item Analysis-›FFT, you will get a pop- up window where you can select the details of the FFT you want to calculate. Once these are set, click OK, and a new graph window will be displayed with the FFT plotted (Figure 5). Once you have had a chance to
 |
Figure 5. FFT analysis is useful in signal processing. |
look at your data, you may have started to form an idea of a model that represents the system you were measuring. An idea is not enough, however. You actually need to do some calculations and see whether your model fits the data you collected. Two options are available. The first is to use the Fit Wizard. You can access it by clicking on the menu item Analysis->Fit Wizard. This pops up a window where you can build a function describing your model. Once you have built up your model, click the button called Fit. This pops up a new window where you can select the details of doing the actual fitting of the generic function to your data. Here you can set the initial guesses and select the algorithm used to do the actual fitting (Figure 6). You also can set how many iterations to try and the tolerance of when you can stop. When everything is set to your satisfaction, click the Fit button. This pops up a new graph window, plus an output window detailing the results of trying to do the fit. The other option of trying to fit your model is to start with your
 |
| Figure 6. The Fit Wizard lets you define your model and see how well it fits your data. |
initial graph. ln this case, you start willing to go through the manual. by right-clicking the main graph window and selecting the Analyze option. This opens a submenu where you can select one of a number of common defaults, such as linear, polynomial, exponential growth or decay, among several others. You also can open the Fit Wizard from this submenu. This article has been only a very short introduction. Lots of other functions are available if you are Also, you have the option of using Python as a scripting language within SciDAVis in order to do even more complicated data analysis. You might need to compile from source, however, if the binaries available to you don't include this functionality. Hopefully, you will take the time to learn what may be a very useful tool for analyzing your experimental data. -Jorzvnznmmn
Aucun commentaire:
Enregistrer un commentaire