Questions about LDA method in Unscrambler?


Please explain why do you think this question should be reported?

Report Cancel

I´m working with LDA in Unsc X ver 10.4 trying to classify samples with NIR spectral information. The LDA results obtain a prediction matrix with the discriminant value for each class. How calculate Unsc these discriminant values?
At the same time, I would like to obtain the discriminant functions to evaluate the variables with more discriminant power. Is it possible in LDA?
Thanks in advance.

About the Author

Thread Reply

  1. External Admin

    Please briefly explain why you feel this answer should be reported .

    Report Cancel

    LDA as implemented in Unscrambler has three options. These are based on an estimation of
    the pooled covariance of the individual classes:
    . Linear
    . Quadratic
    . Mahalanobis distance.

    For the Linear and Quadratic option, prior probability may be given in as a parameter.

    The LDA algorithm is given below:

    1. Estimate the mean for all classes

    2. Estimate the posterior log probability (for Linear and Quadratic models only) for each
    sample and class:
    Distance = log( prior(k) ) – ½ ((xi – xmk)^2 / R)
    , where xi is the vector of variables for each sample, xm is the class mean, k is the class index, and
    R is the pooled estimate of the covariance matrix of X.

    3. Assign the sample to the class with the smallest Distance

    The distances and the class assignment are given out in the table named Prediction in the
    Results node.

    We regret the discriminant functions themselves are not returned.

Leave an answer