標籤: PLS

發佈於

【張偉豪專欄】PLS-DA的用途

請問老師PLS-DA的用途是?

張偉豪老師回覆:

PLS Discriminant Analysis (PLS-DA) 看來與Finite Mixture Partial Least Squares (FIMIX-PLS)類似,均是作為分群
但看來又不太一樣
而且PLS-DA有特定的軟體
所以您應該參考 XLSTAT statistical software
其中有提供PLS-DA操作

發佈於

【張偉豪專欄】PROCESS的能與不能?

1. When we run the process for mediation or moderation, how to calculate the value of latent variable(潛在變數)? do we calculate the mean of total measure items. For example, attitude has 3 measure items then we use the mean of 3 measures for attitude construct?

2. Process is also applicable for the construct which has formative indicators(形成型指標)?
3. If my latent construct is 2nd order construct, how to calculate the value of this latent construct? Again, should I get the mean of total of first order constructs? For example, experiential marketing has 4 first order constructs, then the experiential marketing construct is 2nd order so how can I get the value of this 2nd order construct in the process?

張偉豪老師回覆:
1.process 只能處理觀察變數,潛在變數一定要用SEM來進行分析
所以在Process分析中要先將構面的題目取均值再進行分析.

2.形成型指標只有PLS可以處理

3.SPSS無法處理二階的問題,因此不論二階構面有多少題,最後仍要平均成1個題目作代表.
只有SEM及PLS可以處理二階以上的模型

發佈於

【張偉豪專欄】PLS需要考慮常態假設的問題嗎？

反映型指標模型評估有提到須考慮Multi-normality，像我的模型因Mardia’s normalized estimate 為 44.001，違反多元常態的假設，所以我不用SEM而用PLS，這是對的嗎？

張偉豪老師解答

在PLS的分析中，可以不用考慮常態假設的問題，這也是PLS的優點之一，所以您是對的。