| T O P I C R E V I E W |
| PhD_Bashar |
Posted - 08/02/2016 : 08:12:38 AM
Origin Ver. and Service Release (Select Help-->About Origin):
Operating System:
I hope that someone can help clarifying some issues related to PCA in Origin.
I am trying to make a predictive model of dissolution release depending on PCA findings of 18 variables which effect the dissolution release rate, using pro Origin, in this context I have some questions?
1- The scores of PCs of each label in X column, should be calculated as the sum of variables data multiplied by the respective extracted eigenvectors (coefficients of PC1, PC2 ...), but I could not get the same scores what Origin showed. Knowing I made the PCA for the original data without standardizing it (subtracting means), but I also didn't get the same scores with multiplying the subtracted means data, should I in general input standardized input data before running PCA using Origin? And how can I calculate the scores correctly?
2-from the correlation matrix of these 18 variables, can I get equations that describe those correlations between variables?
3- I calculated the value of each variable in PC1 for example, by dividing the variable’s extracted eigenvector by the sum of coefficients of PC1 then multiplied by 100, is this correct?
4- Could you please refer me an example or explanation for using PCA for making a predictive model?
Thanks for your help in advance.
|
| 1 L A T E S T R E P L I E S (Newest First) |
| Sam Fang |
Posted - 08/03/2016 : 05:16:25 AM
1. If Standardize Scores option is not chosen, scores in each column will be multiplied by eigenvalues. And means will be subtracted if you want to calculate score manually. You can refer to the document for PCA's algorithm.
2. Correlation is only used to describe relation between two variables.
3. We seldom calculated it in PCA. There are several quantities which can calculate it in Partial Least Squares tool (Select Statistics: Multivariate Analysis: Partial Least Squares from Origin menu).
4. Current PCA tool didn't support to predict model. Partial Least Squares tool can.
Sam
OriginLab Technical Services |
|
|
