QUANTILE REGRESSION

QUANTILE REGRESSION 在 自訂表格及進階統計量中可用。

指令會呼叫「分位數迴歸」程序。 需要單一數值應變數。 BY 及 WITH 分別後接因素及共變數的選用清單。 請注意，需要截距項或至少一個預測值才能執行分析。

QUANTILE REGRESSION dependent_varname [BY factor_list] [WITH covariate_list]
[ /MISSING
     [ CLASSMISSING = {EXCLUDE**} {INCLUDE}]]
[ /CRITERIA
     [ QUANTILE = {0.5**} {[values] [value1 TO value2 BY value3]} ]   @@
     [ IID = {TRUE**} {FALSE} ]
     [ BANDWIDTH = {BOFINGER**} {HALL_SHEATHER} ]
     [ CILEVEL = {95**} {value} ]
     [ METHOD = {AUTO**} {SIMPLEX} {INTERIOR_POINT} ]
     [ TOL = {1E-12**} {value} ]   ## 
     [ CONV = {1E-6**} {value}]   $$ 
     [ MAXITER = {2000**} {integer} ] ]
[ /MODEL
     [ effect_1 effect_2 ... ]
     [ REGWGT = weight_varname ]   %%
     [ INTERCEPT = {TRUE**} {FALSE} ] ]  ^^ 
[ /SAVE
     [ PRED([ {varname} {rootname}) ] ]
     [ RESID( {varname} {rootname}) ]
     [ CIPREDL({varname} {rootname}) ]
     [ CIPREDU({varname} {rootname}) ] ]
[ /PREDICT_EFFECTS
     [ NUM_TOP_EFFECTS = {3**} {integer}] ]
[ /PREDICTION_TABLES ]
     [ effect_1 effect_2 ... ]   && 
[ /PREDICTION_LINES ]
     [ effect_1 effect_2 ... ]   ⊕ 
[ /PRINT
     [ PARAMETER** ]
     [ COVB ]
     [ CORB ] ]
[ /PLOT
     [ MAX_CATEGORIES = {10**} {integer} ]   ⊖
     [ PREDICTED_BY_OBSERVED = {FALSE**} {TRUE} ]
     [ PARAM_EST = {50**} {integer} {ALL} ] ]
[ /OUTFILE
     [ COVB = 'savfile' | 'dataset' ]
     [ CORB = 'savfile' | 'dataset' ]
     [ FILE_SEPARATE = {FALSE**} {TRUE} ]
     [ {MODEL = 'filename'} {PARAMETER = 'filename'} ] ]
[ /EXTERNAL ]

**如果省略了次指令或關鍵字，則為預設值。

@ @ 使用唯一值 ∈ [0， 1]

## 針對矩陣操作

$$ 對於數值方法

%% 必須是數值變數

^ ^ 截距或至少一個解釋變數

&& A subset of the effects contained in the model

模型中包含的效果子集

為混合效果繪製的種類數上限

發行歷程