.-
help for ^alpha2^                                            (manual:  ^[R] alpha^)
                                     Adopted by Jeroen Weesie/ICS   Jan 3, 1997
.-

Cronbach's alpha                                           [STB-35: sg66]
----------------

        ^alpha2^ varlist [^if^ exp] [^in^ range] [^, a^sis ^c^oo ^d^etail ^g^en^(^newvar^)^ 
                                ^i^tem ^l^abel ^m^in^(^#^) r^everse^(^varlist^)^ ^s^td ]

See help @matsize@ for instructions on resetting ^alpha^'s limits.


Description
-----------

Like ^alpha^, ^alpha2^ computes the average interitem correlations or covariances 
(aiic) for all pairs of variables (items) in varlist and Cronbach's alpha 
statistic for the additive scale (the "test") formed from these items. In 
addition, ^alpha2^ can be used to obtain
  -- aiic and alpha for all "rest scales" obtained by removing one item; 
  -- the item - test correlations; 
  -- the item - rest (rest = test - item) correlations. 
To facilitate the interpretating of output on typical item-batteries, variable 
labels are optionally included. 


Options
-------

^asis^ indicates that the sense (sign) of each item should be taken as presented
    in the data.  The default is to empirically determine the sense (using 
    principal components factor analysis) and reverse the scorings for any that
    enter negatively. Note that subscales use the same directions of the items.
    Use ^reverse^ (see below) to manually specify the signs (directions) of the 
    items without transforming the data.

^coo^ specifies that cases that missing values should be deleted ("listwise"). 
    The default is pairwise computation of covariances/correlations. 

^detail^ specifies that the matrix of interitem covariance or correlation (if 
    ^std^ is specified). If different numbers of observations have been used 
    to compute correlations or covariance, these are displayed as well.

^generate(^newvar^)^ specifies that the scale constructed from varlist is to be
    stored in newvar.  Unless ^asis^ is specified, the sense of items entering
    negatively is automatically reversed.  If ^std^ is also specified, the scale
    is constructed using standardized (mean 0, variance 1) values of the indi-
    vidual items.  Unlike most Stata commands, ^generate()^ does not employ case-
    wise deletion.  A score is created for every observation for which there is
    a response to at least one item (one variable in varlist is not missing).
    The additive score is divided by the number of items over which the sum
    is calculated.

^item^ specifies that item-test and item-rest correlations and the effects of 
    removing an item from the scale are displayed. 

^label^ requests that the detailed output table is displayed in a compact format 
    that enables the inclusion of variable labels.

^min(^#^)^ that only cases with at least # observations are included in the 
    computations. ^coo^ is a shorthand to ^min(^k^)^, with k the number of variables
    in varlist.

^reverse(^varlist^)^ specifies that the signs (directions) of the variables 
    (items) in varlist should be reversed. 

^std^ specifies that the items in the scale are to be standardized (mean 0,
    variance 1) before summing.


Examples
--------

    . ^alpha2 q1-q10^                        (minimal output: aii-cov and alpha)

    . ^alpha2 bg2cost1-bg2cost6 dotest1-dotest6, std d^      (standardized test)

    . ^alpha2 bg2cost1-bg2cost6 , std g(score)^       (save test scale in score)

    . ^alpha2 q1-q10, item^              (remove individual items one-at-a-time)

    . ^alpha2 q1-q10, i label^             (compatc output with variable labels)


Saved results
-------------

Like ^alpha^, ^alpha2^ returns the average interitem covariance/correlation 
in ^S_4^, the number of items in ^S_5^, and the alpha statistic in macro ^S_6^.


Also see
--------

    STB:  sg66 (STB-35)
 Manual:  ^[R] alpha^
On-line:  help for @factor@