Maximum data size

Q. What is the maximum amount of data that MVSP can handle?

A. The theoretical limits are 2 billion variables and 2 billion cases. The actual amount of data that you can analyze depends on the amount of memory available to Windows (this includes both RAM and hard disk swap file size). Also, different analytical procedures require different amounts of memory for intermediate results. With PCA and CA/CCA most computers can easily analyse thousands of cases, so long as the number of variables is of a lower order of magnitude (in the tens or low hundreds). Clustering and PCO have higher memory requirements, since they must calculate symmetrical distance matrices with the same number of rows and columns as the number of cases. For example, a cluster analysis of a data file with 2000 cases would require a symmetrical matrix of 32 megabytes (2000 x 2000 x 8 bytes per datum). A computer with 96 Mb of RAM will be able to analyze this file easily. A 32 Mb machine, if it has enough disk space for a suitable swap file, will be able to analyze this, but will spend much of its time writing data to and from the disk, slowing down the analysis. The best way to determine if your computer can analyse a data set of a certain size is to simply try it. If MVSP can't get the amount of memory it needs it will give a message saying there is not enough memory.

Labeling scatter plot points

Q. Is there a way to label each point on a scatterplot with the corresponding label from the data file?

A. Yes. 1. Choose the Graphs|Edit Graph menu option and go to the Labels page. 2. In the Data Labels section tick the box marked "On". 3. Press OK. Note that you can also display the label for individual points (one at a time) by clicking on the point with the mouse.

Dendrogram labels too small

Q. I've done a cluster analysis of a large number of cases. However, the labels on the dendrogram are so small that I can't read them. What can I do?

A. You can zoom in on the graph using the Graphs|Zoom In menu item to look at the labels on the screen. You can also tick the "Text dendrogram" box on the Advanced page of the clustering dialog. This will create a text-based dendrogram in the results window, with normal sized fonts, that you can study to see the objects. The size of the labels on the dendrogram can also be adjusted in the Options|Font (or Graphs|Edit Graph) dialog box.

PCA scatter plot points not spread out

Q. The points on my PCA scatter plot are all clustered around the center of the second axis, rather than spread out. Why?

A. Ordination results are, by default, plotted with the same X and Y scales, so that the dispersion of points along the two axes is comparable. As a result, the graphs are best displayed and printed as square graphs, rather than rectangular. In your plot the amount of variance accounted for by the second axis is probably much less than for the first one, so the distribution of the points reflects this. If you wish to use different scales on the two axes, so that the data points are spread out as much as possible, you can do this through the Axis page on the Edit Graph dialog box: 1. Choose the Graphs|Edit Graph menu option and go to the Axis page. 2. In the Apply to Axis box select the X option. 3. In the Scale section choose either zero or variable origin (depending on whether you want to force the scale to include zero). 4. Repeat step 2, choosing Y Primary instead, then repeat step 3. 5. Press OK.

MVSP CCA results differ from Canoco

Q. When I compare the CA/CCA results of MVSP with those of CANOCO (using the same scaling) many of the numbers are different beyond the third or fourth decimal place. Why is this?

A. Very early versions of CANOCO (v. 3.12 and earlier) used a fairly lax criterion for determining when to stop calculations. This has been pointed out in an article by Oksanen & Minchin (1997, J. Vegetation Sci. 8, 447-454). If you set the accuracy of a MVSP analysis (on the Advanced tab of the CA dialog) to a low level such as 1E-5, you will find that the results are identical to CANOCO. Later DOS versions of CANOCO, and the current Windows version, do not have this problem.

Drawing graphs produces errors; used to work OK

Q. Ever since I recently experimented with some of the graph customization options all attempts at creating a graph have failed with various error messages. This only happens with one graph type. What's gone wrong and how can I fix it?

A. When you make changes to a graph the settings are saved, so that you can create your own style of graphs and have it automatically used for future graphs. However, certain combinations of graph customization settings seem to create problems for drawing new graphs. This also only seems to happen on certain computers. The easiest way to fix this is the use the Graphs|Reset Defaults menu option. This will erase all graph customizations and redraw all current graphs with the default settings. This will not affect any graphs you have saved to desktop files.

Problems creating MVSP files manually

Q. I'm trying to automatically create MVSP files from my own program but MVSP has problems reading them. What am I doing wrong?

A. The most common reasons for those errors you've run across are:

  1. The number of rows and columns specified in the header do not match the actual number of data rows and columns. The labels *do not* count as a row or column, these numbers should reflect the data only.
  2. Either the row or column labels have been omitted; both must be present, with the column labels preceeding the entire data set and the row labels at the start of each row of data.
  3. One or more of the labels or data values has a space in the middle, so that MVSP reads it as two labels or values.
  4. All zero values must be included, not left blank.
  5. The data values must be separated by spaces, not tabs or commas.
  6. If none of these work e-mail one of the files to us and we'll have a look at it to see what is wrong.  

I get an "Action canceled!" error when trying to view the help file

Q. When I try to view the help file I get a message saying "Action canceled! Internet Explorer was unable to link to the web page you requested". How can I get it to work?

A. This is caused by a security update released by Microsoft in 1995. Basically, this stops compiled HTML help files from being displayed if they are run from the Internet or over a network. Help files installed on your local hard disk will still work OK.

If you run MVSP once with administrator privilages then this should fix the problem. If not then you can fix this by having the administrator of your computer make some changes to the Windows registry. The changes to be made are described at the Microsoft Knowledge base, article 896358. There is also a free utility called HHReg that automates the process.

About KCS

Kovach Computing Services (KCS) was founded in 1993 by Dr. Warren Kovach. The company specializes in the development and marketing of inexpensive and easy-to-use statistical software for scientists, as well as in data analysis consulting.

Mailing list Join our mailing list

Home | Order | MVSP | Oriana | QDA Miner
Stats Books | Stats Links | Anglesey


Like us on Facebook Facebook

Get in Touch

  • Email:
  • Address:
    85 Nant y Felin
    Pentraeth, Isle of Anglesey
    LL75 8UY
    United Kingdom