LOADM <matrix_file>,<format>,L
works in principle like MAT LOAD but writes the matrix in the edit
field (and in the output file) in block form where the width of each
block is limited by the visible part of the edit field.
The number and order of rows and columns to be written can be selected
by ROWS and COLUMNS specifications. For example, ROWS=3,2,1,4 informs
that rows 3,2,1,4 only and in this order will be written.
To emphasize exceptional elements of the matrix (like high loadings
of a factor matrix) in the display, specifications LIMITS and SHADOWS
can be used.
Another alternative is to use the SHADOW_MATRIX specification as shown in
LIMITS gives a list of upper bounds of various classes and SHADOWS
gives a list of corresponding shadow values.
Example: LIMITS=-0.7,-0.3,0.3,0.7,1
SHADOWS=7,1,0,1,7 (0=normal display)
Another alternative is to use the SHADOW_MATRIX specification as shown in
Formats of elements in LOADM:
Parameter <format> is given in one of the following ways:
1. LOADM <matrix_file>,12.1234,L
writes elements with 2 places in the integer part and with 4 decimals.
2. LOADM <matrix_file>,(10),L
gives elements with 10 characters according to the maximum element
by absolute value. Thus when the greatest value is 1998.261213,
the selected format corresponds to 1234.1234 (if negative values
appear in the matrix) or to 1234.12345 (if not).
3. LOADM <matrix_file>,(C10),L
works as the previous alternative but the format is selected
separately for each column of the matrix.
Extended form of LOADM especially for factor matrices
according to the style suggested by Lauri Tarkkonen:
LOADM <matrix_file>,<format>,L
with specifications LIMITS, SHADOWS, SORT, COLUMNS, SUMS, POSDIR, WIDE
writes a (rotated factor) matrix in the edit field by
1. sorting (SORT) the rows hierarchically by factors and loadings
of variables,
2. inserting long descriptive names of variables at the end of
each row (provided that a name of data giving those names is
indicated by SORT),
3. displaying the elements with different shadings according to their
sizes (LIMITS, SHADOWS as described earlier, see LOADM?),
4. sorting the columns according to the sums of squares (COLUMNS=SORT),
5. displaying marginal sums of squares (SUMS=2) or sums (SUMS=1),
6. orienting the factors (signs of loadings) according to the same
rules as in the POSDIR operation (POSDIR=1,2, or 3), see POSDIR?,
7. allowing wide matrices to be written as a single block (WIDE=1)
if the line length is large enough.
The most important specification is SORT of the form
SORT=1
or
SORT=<name_of_data_file>
In both cases the variables are written in the order of their
loadings in different factors as follows. At first the factors
are studied according the order of the sums of squares of loadings.
Variables having their highest loadings in the strongest factor
are sorted to descending order according to the absolute
values of their loadings in that factor. These variables in this
order will appear as first ones in the output. The same procedure
is applied to the remaining variables according to the second
strongest factor and so on.
In the latter form of SORT the long names of variables are taken
from the given data file and appended to the end of each row.
This helps interpretation of factors.
SORT=-1 or SORT=-<name_of_data_file>
leads to the same sorting procedure but now within each factor
the loadings are sorted by their signed values.
An extra parameter in all forms of SORT, for example, SORT=1,0.3
excludes variables with a highest loading less than a given value (0.3)
from the sorting procedure and puts those variables (rows) at the end
of the matrix.
Sucro /LOADFACT <data_file>,<rotated_factor_matrix>
does the job automatically with selected values of extra specifications
and gives a ready-made work scheme which may be improved by the user.
See
/LOADFACT ?
Help System (web edition)