Stata 11 help for candisc help candisc dialog: candisc also see: discrim lda ------------------------------------------------------------------------------- Title [MV] candisc -- Canonical linear discriminant analysis Syntax candisc varlist [if] [in] [weight], group(groupvar) [options] options description ------------------------------------------------------------------------- Model * group(groupvar) variable specifying the groups priors(priors) group prior probabilities ties(ties) how ties in classification are to be handled Reporting notable suppress resubstitution classification table lootable display leave-one-out classification table nostats suppress display of canonical statistics nocoef suppress display of standardized canonical discriminant function coefficients nostruct suppress display of canonical structure matrix nomeans suppress display of group means on canonical variables ------------------------------------------------------------------------- priors description ------------------------------------------------------------------------- equal equal prior probabilities; the default proportional group-size-proportional prior probabilities matname row or column vector containing the group prior probabilities matrix_exp matrix expression providing a row or column vector of the group prior probabilities ------------------------------------------------------------------------- ties description ------------------------------------------------------------------------- missing ties in group classification produce missing values; the default random ties in group classification are broken randomly first ties in group classification are set to the first tied group ------------------------------------------------------------------------- *group() is required. statsby and xi are allowed; see prefix. fweights are allowed; see weight. See [MV] discrim lda postestimation for features available after estimation. Menu Statistics > Multivariate analysis > Discriminant analysis > Canonical linear discriminant analysis Description candisc performs canonical linear discriminant analysis (LDA). What is computed is the same as with [MV] discrim lda. The difference is in what is presented. See [MV] discrim for other discrimination commands. Options +-------+ ----+ Model +------------------------------------------------------------ group(groupvar) is required and specifies the name of the grouping variable. groupvar must be a numeric variable. priors(priors) specifies the prior probabilities for group membership. The following priors are allowed: priors(equal) specifies equal prior probabilities. This is the default. priors(proportional) specifies group-size-proportional prior probabilities. priors(matname) specifies a row or column vector containing the group prior probabilities. priors(matrix_exp) specifies a matrix expression providing a row or column vector of the group prior probabilities. ties(ties) specifies how ties in group classification will be handled. The following ties are allowed: ties(missing) specifies that ties in group classification produce missing values. This is the default. ties(random) specifies that ties in group classification are broken randomly. ties(first) specifies that ties in group classification are set to the first tied group. +-----------+ ----+ Reporting +-------------------------------------------------------- notable suppresses the computation and display of the resubstitution classification table. lootable displays the leave-one-out classification table. nostats suppresses the display of the table of canonical statistics. nocoef suppresses the display of the standardized canonical discriminant function coefficients. nostruct suppresses the display of the canonical structure matrix. nomeans suppresses the display of group means on canonical variables. Examples Setup . webuse rootstock Fit linear discriminant analysis (LDA) model with equal prior probabilities for the six rootstock groups and display the canonical results . candisc y1 y2 y3 y4, group(rootstock) Fit the same model, but use prior probabilities of 0.2 for the first four rootstocks and 0.1 for the last two rootstocks and display only the canonical discriminant test statistics and the leave-one-out classification table . candisc y*, group(rootstock) priors(.2,.2,.2,.2,.1,.1) lootable notable nocoef nostruct nomeans Replay results allowing the default display . candisc Saved results candisc saves the same items in e() as [MV] discrim lda with the exception that e(subcmd) is not set and the following e() results are different: Macros e(cmd) candisc e(title) Canonical linear discriminant analysis Also see Manual: [MV] candisc Help: [MV] discrim lda postestimation, [MV] discrim lda