Optifact - v1.2 - 11 Mar 2008

Find the most appropriate summated rating scales from a list of items


optifact varlist [if] [in] [weight] [, options]

Options top(#) The number of the best candidate scales to list after analysis (default 10)

smallest(#) The smallest number of items allowed in the candidate scales (default 3)

largest(#) The largest number of items allowed in the candidate scales (default is all the items)

maxfact(#) The maximum number of factors (latent concepts) in the scale (default is 1)

minslope(#) The minimum allowable slope in the canonical correlation model (default .01)

criteria(varlist) A list of variables for criterion validity testing using canonical correlation

type(string) The type of factor analysis to use. Must be one of pf pcf or ml. Default is pf. See help for factor

sig Output the statistical significance of the model (assumes ml models). Note this is the probability from the factor model (with ml option) that all parameters of the model are zero in the population. To get the statistical significance of alpha, bootstrap the particular scale you are interested in. For example: bootstrap "factor item1 tem2 tem4 item6" r(alpha), reps(400)

by may be used with optifact; see by.

aweights and fweights are allowed; see weight.

Note: you must first install matsort before you can use optifact


optifact examines every possible combination of items specified by varlist then lists out the scales with highest alpha and the given number of factors that can be constructed from these items.


. sysuse auto

. optifact price mpg rep78 headroom trunk weight length turn displacement gear_ratio, largest(4) criteria(foreign)


Paul Millar www.ucalgary.ca/~pemillar pemillar@ucalgary.ca