I will start a collection here of some frequent questions about Minitab. Again, I would like to stress more general things. If you have some suggestions for questions that I can post answers to here ... shoot me a note at ....Dennis Roberts

Click on the link to the FAQ you want an answer to

  1. Ranking data correctly in Minitab
  2. Controlling decimal places in Minitab
  3. Reversing rows and columns
  4. Making a plot smaller
  5. Lose data with single column commands
How do I get Mintab to rank data correctly?
Ever need to rank some data? Well, in some cases, like Grade Point Averages, we would assign a rank of 1 to the highest or largest GPA value. However, in other cases, like professional averages for strokes per rounds over the year, a rank of 1 would be assigned to the golfer with the LOWEST score. Thus, somethimes big numbers should have high ranks and other times, low ranks.

There is a command in Minitab called RANK. It works in the following way. When the data points are lined up in order, from high value to low value, then Minitab will assign the LOWEST rank value (=1) to the LOWEST data value, and the HIGHEST rank value to the HIGHEST data value. So, if the numbers are 34, 23, 19, and 10 ... and the data are in c1, then using the MTB> rank c1, c2 command will put the rank values of 4, 3, 2 and 1 in c2. But, in many cases, high data values mean good or the best and, those are the values you want to assign the ranks of 1 or 2 or 2 to ... and the big rank values would be assigned to the LOW data values; ie, just the OPPOSITE to what Minitab does. How can I get Minitab to 'cooperate'? Well, again, say that the data that you want to rank are in c1. If you use a LET command like the following, you can get Minitab to REVERSE the way it ranks values. Look at this:

MTB> let c2=(n(c1)-rank(c1)+1))

If you did that with the data 34, 23, 19, and 10 ... and printed out c2 ... the ranks would then be 1, 2, 3, and 4 respectively.

How can I get Minitab to give me more control over decimal placement?

First, it is important to understand that the actual calculations for work done in Minitab carry many, many places ... even though you don't see them in output. Thus, if you see a correlation value of .456 ... that is accurate to that many places ... even if you had taken it 10 more places out and had rounded BACK to 3. So, keep that in mind.

Well, the bottom line is that you do NOT have much control over decimal placement. Minitab uses some internal algorithm for 'deciding' on this ... and what you normally get depends on the variables you have, and some compromise for decimal places. Are you stuck therefore? Is there nothing you can do?

Essentially YES but, here are a couple of suggestions. If you have some data in columns and want to DESCRIBE the data .. there is a subcommand called DECIMAL that will allow you to have some control over the DESCRIBE output. So, if you had some data in c1-c3, and then did:

MTB> desc c1-c3
SUBC> deci 3.

The DESC output would have 3 decimals places. Sometimes however, if you get too greedy, it will list it in scientific notation ... now THAT's a blast. However, you can get a little more decimals in DESC if you want.

What if you were doing some correlation problems and had several variables? You could use the MTB> corr c1-c3 command and you would get some decimals in your output. But, if you stored the correlation output in a MATRIX ... and then printed the matix, you will get more decimals. Look at the following.

MTB > corr c1-c3

             C1       C2

C2       -0.266

C3        0.616   -0.150

MTB > corr c1-c3 m1

MTB > prin m1

 Matrix M1


  1.00000  -0.26561   0.61617 

 -0.26561   1.00000  -0.15000 

  0.61617  -0.15000   1.00000 

MTB > 

For example, the r between c1 and c2 is -.266 in its usual output form but, if you store the values in a matrix, we would add a couple of decimals to that (-.26561). I am not suggesting that you need that many places, I am only saying that IF you want more, you could get more places this way.

Now, you can get Minitab in the DATA WINDOW to display more decimals although this will NOT impact on the output in any analysis you do and have displayed in the session window. For example, if you are in the data window, you can highlight any cell in the column you want to change the decimal places .. say from 3 to 5 ... by RIGHT CLICKING on the mouse ... selecting format column, then numeric, then find the Format ... fixed column width and put in the box whatever you want to change it to ... and click on OK. You will see the new decimal place arrangement automatically applied to all values in the column. Now, you will get the new decimal place arrangement if you print out the worksheet but, as I said, it will NOT impact any on any analysis you do and the output for it that you see.

How can I get Minitab to reverse rows for columns, and columns for rows?
Now and then, you might have access to or have imported some spreadsheet or worksheet that, unfortunately to you, has the columns and rows backwards from what you would like. For example, in doing gradebooks in classes, we normally would want the individual students on the side axis and, the scores on tests and projects across the top axis. The reason for this normally is that most packages give you more options and routines for working with the columns than the rows. For the above example, it would be easier to find things like means and standard deviations of test and project scores if these were the columns of information.

So, what if I had a very small set of data ... 5 rows and 4 columns ... where the columns are the people and, the rows are the characteristics (variables) on the people. Don't pay ANY attention to the numbers here ... I just made them up. So, for person 1, he/she got 18 on variable 1, 16 on variable 2, and so on. What if we want to rearrange the data table ... so that the FIRST LINE is for person 1 ... and would have the values of 18, 16, etc. ACROSS the first row ... and not down the first column? Look at what I have done to accomplish this.

 Row  person1  person2  person3  person4

   1       18           50           0          263

   2       16           50           1          200

   3       18           45           1          276

   4       20           31           1          202

   5       17           42           1          244

MTB > copy c1-c4 m1

=================> You don't need to do the command below

MTB > prin m1

Data Display

 Matrix M1


   18    50     0   263 

   16    50     1   200 

   18    45     1   276 

   20    31     1   202 

   17    42     1   244 

================> You do need to do the following

MTB> transpose m1 m2

================> You don't have to do the command below

MTB > prin m2

Data Display

 Matrix M2


   18    16    18    20    17 

   50    50    45    31    42 

    0     1       1      1     1 

  263   200   276   202   244 

================> You do have to do the command below

MTB > copy m2 c5-c9

MTB > prin c5-c9

Data Display

 Row     C5     C6     C7     C8     C9

   1        18     16     18     20     17

   2        50     50     45     31     42

   3          0      1       1       1      1

   4        263   200    276    202    244

Sorry about some of the lining up in the columns ... the formatting is not perfect in my html file. On the Minitab output screen though, it would be.

Here are the essential commands in this.

MTB > copy c1-c4 m1
MTB> transpose m1 m2
MTB > copy m2 c5-c9

The first copy command transfers the data in c1-c4 into a matrix that will be 4 wide and 5 deep. The transpose command will take the data in m1 ... and turn it 90 degrees around ... and call that new matix m2. The second copy command will now give NEW column identifiers to the top side of m2 ... and in the data display above of prin c5-c9, you see the result of the reversed rows and columns. Of course, though I did not do it above, you would need to give names to the new columns you have created. A handy little trick indeed!

How can I get Minitab to make smaller graphs than are usually printed in the graph window?
On some occasions, you might want to reduce the size of a graph .... here is how you can do it. Keep in mind however that severe reductions in graph size will tend to blur the graph ... so, large reductions are not recommended.

Let's say that we have some X and Y data and make a plot. The default plot looks like the following::

plotreg.jpg - 10691 Bytes

Now, what if you would like to modify this graph, mainly to make it smaller. If you pull down the GRAPH window to PLOT ... click on that ... you will see a dialog box. Look at the following.

dialog1.jpg - 44420 Bytes

In the dialog box, you would have clicked on the c1 and c2 variables to list them as X or Y. Then, at the bottom of the box, click on REGIONS .... then select FIGURE. This would give you another dialog box ... see the following.

dialogg2.jpg - 37629 Bytes

Note that there are X and Y min and max boxes. Here, you enter your relative change in size ... for both X and Y ... and note that the default is 0 and 1. Now, I could have only changed the MAX on X and Y to say ... .8 of the max size, but the graph that came out would be more to the lower left corner of the graph window. So, what I wanted to do is to reduce the size but keep the graph in the center. Note that I put .2 and .8 (moved over each way the same amount) for both X and Y. See what the new smaller graph looks like.

plotmod.jpg - 9561 Bytes

While the graph is smaller, it is less clear too. Keep that in mind when reducing the size of graphs. NOTE: You cannot make the graphs larger since, values larger than 1 will not be accepted in the MAX boxes.

Why do I lose data in a column when I use commands like ... mean c1 c2, or stdev c5 c6?
This is indeed a very good question. Now, early on in Minitab's life, there was an emphasis on doing calculations on a single column of data ... like finding the mean or standard deviation for a column. Thus, there were created a number of SINGLE COLUMN commands. Look at the HELP for one of these ...

Command Syntax

MEAN of the values in C [put into K]

Calculates the arithmetic mean of all the values in a column and optionally stores the result. Missing values are omitted from the calculation.

Now, note that the syntax is to find the mean for data in some C ... BUT, you have an option for storing that result in some other location ... like a K location OR ... even in a column. Also note that you do NOT have the luxury of finding the means for multiple columns all at once with this single column command.

So, say you have some data in columns c1 and c2 ... and then want the means for both ... and do the following:

MTB> mean c1 c2

Well, Minitab will find the mean for the data in c1 but, since there is an option for storing the data ... and you have put c2 second on the line ... it will ERASE ANY DATA IN C2 ... and then store the mean for the data in c1 as the FIRST ROW ENTRY in c2 ... thus, what you would have now is 1 value in c2 ... and it would be the mean for c1. Thus, the data that you had in c2 would be gone ... much to your surprise.

There are a number of these single column commands ... here is a partial list ...

Keep in mind that for all of these commands, you can work on ONE column at a time ... and have an OPTION to store the result in some other place ... like a k storage location or even in a column. While these commands can come in handy now and then, you MUST be careful when using them so as not to erase data in columns unexpectedly.
Back to the Top of the Page ... Top of Page ... MTB Resources Page ... MTB Resources Page