Visualpls手冊以及學(xué)習(xí)筆記.docx

Visualpls可以建立二階因子模型，但是只能做PLS演算，BootStrap和JackKnife算法不支持二階變量，所以R方和T值都無法得到，怎么對二階因子模型進行評估呢？或者有沒有其他的PLS軟甲你可以做二階因子模型的評估？AMOS直接從SPSS導(dǎo)入數(shù)據(jù)，但是SmartPLS和VisualPLS要求.txt或.dat后綴的數(shù)據(jù)，小白求各位高手科普，該如何建立SmartPLS和VisualPLS的數(shù)據(jù)源1. 可以先把數(shù)據(jù)輸入到excel里面，點另存就行了。2. 用SPSS打開數(shù)據(jù)后，點“另存為”，在保持格式里選“用逗號隔開 .csv就好了信度檢驗：內(nèi)部一致性信度【信度，一般要求信度在0.7以上，可放松到0.6】、合成信度【各組樣本平均值之間的差異顯著性檢驗，反映的是潛變量內(nèi)部指標(biāo)的一致性，一般采用CR指標(biāo)（VISUAL-PLS），一般要求大于0.7】效度檢驗：一致性效度、判別效度【檢驗概念之間的差異程度，主要指標(biāo)是AVE平方根與潛變量間的相關(guān)系數(shù)（VISULA-PLS），一般而言，若一個潛變量的AVE值大于相關(guān)系數(shù)，說明效度存在沒差異明顯】、內(nèi)斂效度【主要是指AVE值，一般認(rèn)為大于0.5即可】路徑系數(shù)檢驗：采用Bootstrap算法【VISUAL-PLS軟件（判定指標(biāo)主要是定標(biāo)決定系數(shù)Rsq，為負(fù)數(shù)路徑不成立）】Creating and Editing PLS ModelA data file must be prepared before the creating of PLS model. The requirement for a legitimate data file was discussed in the previous section. Then the following steps are detailed in squence:1.Click File|New Model or click button to create a new PLS Project. The default file type is vpl. PLS code will be written into this file. You need to finish these steps to create a valid PLS code file:Step 1: Specifying PLS Project File Name.Step 2: Locating data file.o The data type is raw data. The default extension name is txt or dat.o Once the PLS project is created, the data file cannot be changed.Step 3: Building the measurement model.All the indicators will be read from the first row in the data file and listed in the manifest variable list box. The indicators will be removed when they are assigned to a latent variable. Step 4: Constructing the structural model.The path model can be created by dragging lines between constructs. Step 5: Specifying model parameters.Open PLS Projectclick the button to open PLS Project.Select CasesSelect Cases provides methods for selecting a subgroup of cases based on criteria that include variables and expressions. Model BuildingAll the indicators will be read from the first row in the data file and listed in the manifest variable list box, shown in the left part of the window.New ConstructConstructs must be created before assigning manifest variables. Three operations are related to managing latent variables. o New Construct: By clicking the Tool/New Construct or button, then click on the canvas in the proper position, you can create a new construct. o Delete Construct: Highlight a latent variable and click Delete to remove it. The associated manifest variables will be released.o Rename: To rename a latent variable, double-click it in the construct. A dialog window will pop up to accept a new name. You need to follow the naming rules described above to add a valid variable name. Assign indicators to constructYou can assign manifest variables after adding new constructs. Just double-click on the construct, the associated Indicators and Constructs window will show up. To assign indicators to this construct, just select the un-used indicators (shown in the left list box) and click the Assign button. Multiple selections are supported. The associated manifest variables will move to the right part. If you want to release indicator (s) associated with a construct, and select the indicator (s) you will release in the right list box, then click Remove button. Click OK button, the manifest variable list box will display the relationship between indicators and constructs. Note: You can specify the direction of the manifest variable, either Reflective or Formative.Formative v.s. Reflective IndicatorsFormative and reflective are represented in both windows, as depicted in the picture below. Constructing the path modelThe path model can be created by drawing arrow-lines between variables. To add paths between constructs, press the button, then click the the independent variables and drag to the dependent variables. To remove a path, just click on the path and press the DEL key , or right-click the mouse button, select Delete Path to remove the path.Specifying model parameters.Default parameters are pre-set and estimate the model with general constraints and produce commonly used results. You can make changes to these parameters by selecting items from the parameter lists, or by filling in parameter boxes. It is recommended not to change them unless you are familiar with the meaning of the parameters. If you have made any changes and want to restore the settings, click Reset. The parameters for a correlation matrix model and a raw data model are different. The following is a list of both sets of parameters. Please read the help file in Lvpls package (pls.hlp) for details.Calculation of Stone-Geisser Q2: to estimate the Q2 coefficients, the type of blindfolding static and Blindfolding omission distance in no. of cases must be selected.Save ModelYou can save the current model by clicking Save Model.Save AsYou can save the current model under another name by clicking Save As.Interactoin Effects and ModeratorsPartial Least Squares Product Indicator ApproachThe Product indicator approach, proposed by Chin(1996), was support in the VisualPLS. The predictor (X) and moderator (Z) indicators are used to create product indicators by creating all possible products from the two sets of indicators. These product terms are used to reflect the latent interaction variable. For example, if the predictor (X) has three indicators and the moderator variable (Y) has 2 indicators, as graphically depicted below, we would have six measures for representing the construct of interaction effect (X*Z).Step to Model Interaction Effects with VisualPLSStep 1: Generate Product Terms for Interaction EffectsStep 2: Create a Construct for Interaction EffectStep 3: Assign product Indicators to ConstructsHow to generate Product Terms for Interaction EffectsClick the Button, then select New product terms (for interaction)If you want to generate product terms using existing construct, then select Constructs button, else select Indicators buttonPress Constructs Buttonthe Generate Product Terms from Constructs window appear. The user then select predictor construct and moderator variable to generate product terms. Be sure to assign 2-character Alias for predictor and moderator var