Skip to main content
light editing
Source Link
amoeba
  • 107.2k
  • 36
  • 321
  • 346

Using Does it make sense to run LDA on several principal components in a linear discriminant analysis for a diagnostic testand not on all variables?

I am interested in building a linear discriminant function to discriminate between 2 groups, out of 60 variables. (I'm planning to select the most discriminative of the variables for a future diagnostic test.) I have calculated the area under the ROC curve for each of these variables individually and none havehas an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals (these are the two groups).

I have tried to reduce the number of variables using principal component analysis (PCA). There are 3 components accounting for 83% of the variation. But unfortunately, all 60 variables have similar weightings (loadings) in the 3 components, so I can't pick just few. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function, but 60 is too many, especially given the small sample.

I wondered if, rather than use the 60 variables, it is possible to use the 3 principal components themselves to buildin a linear discriminant function? I am using Stata for the analysis. Does anyone have any suggestions (LDA)?

Using principal components in a linear discriminant analysis for a diagnostic test

I am interested in building a linear discriminant function to discriminate between 2 groups, out of 60 variables. (I'm planning to select the most discriminative of the variables for a future diagnostic test.) I have calculated the area under the ROC curve for each of these variables individually and none have an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals (these are the two groups).

I have tried to reduce the number of variables using principal component analysis. There are 3 components accounting for 83% of the variation. But unfortunately, all 60 variables have similar weightings (loadings) in the 3 components, so I can't pick just few. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function, but 60 is too many, especially given the small sample.

I wondered if, rather than use the 60 variables, it is possible to use the 3 principal components themselves to build a linear discriminant function? I am using Stata for the analysis. Does anyone have any suggestions?

Does it make sense to run LDA on several principal components and not on all variables?

I am interested in building a linear discriminant function to discriminate between 2 groups, out of 60 variables. (I'm planning to select the most discriminative of the variables for a future diagnostic test.) I have calculated the area under the ROC curve for each of these variables individually and none has an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals (these are the two groups).

I have tried to reduce the number of variables using principal component analysis (PCA). There are 3 components accounting for 83% of the variation. But unfortunately, all 60 variables have similar weightings (loadings) in the 3 components, so I can't pick just few. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function, but 60 is too many, especially given the small sample.

I wondered if, rather than use the 60 variables, it is possible to use the 3 principal components themselves in a linear discriminant analysis (LDA)?

added 231 characters in body; edited tags
Source Link
ttnphns
  • 58.8k
  • 53
  • 287
  • 512

I am interested in building a linear discriminant function using the bestto discriminate between 2 groups, out of 60 variables. (I'm planning to select the most discriminative of the variables for a future diagnostic test.) I have calculated the area under the ROC curve for each of these variables individually and none have an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals (these are the two groups).

I have tried to reduce the number of variables using principal component analysis. There are 3 components accounting for 83% of the variation. UnfortunatelyBut unfortunately, all 60 variables have similar weightings (loadings) in the 3 components, so I can't pick just few. I I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function, but 60 is too many, especially given the small sample.

I wondered if, rather thanrather than use the 60 variables, it is possible to use the 3 principal componentsprincipal components themselves to build an LDFa linear discriminant function? I am using Stata for the analysis. Does Does anyone have any suggestions?

I am interested in building a linear discriminant function using the best of 60 variables for a diagnostic test. I have calculated the area under the ROC curve for each of these individually and none have an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals.

I have tried to reduce the variables using principal component analysis. There are 3 components accounting for 83% of variation. Unfortunately all 60 variables have similar weightings in the 3 components. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function but 60 is too many, especially given the small sample.

I wondered if, rather than use the 60 variables, it is possible to use the 3 principal components to build an LDF? I am using Stata for the analysis. Does anyone have any suggestions?

I am interested in building a linear discriminant function to discriminate between 2 groups, out of 60 variables. (I'm planning to select the most discriminative of the variables for a future diagnostic test.) I have calculated the area under the ROC curve for each of these variables individually and none have an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals (these are the two groups).

I have tried to reduce the number of variables using principal component analysis. There are 3 components accounting for 83% of the variation. But unfortunately, all 60 variables have similar weightings (loadings) in the 3 components, so I can't pick just few. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function, but 60 is too many, especially given the small sample.

I wondered if, rather than use the 60 variables, it is possible to use the 3 principal components themselves to build a linear discriminant function? I am using Stata for the analysis. Does anyone have any suggestions?

deleted 1 characters in body
Source Link
Nick Cox
  • 59.5k
  • 8
  • 136
  • 212

I am interested in building a linear discriminant function using the best of 60 variables for a diagnostic test. I have calculated the area under the ROC curve for each of these individually and none have an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals. I

I have tried to reduce the variables using principal component analysis. There are 3 components accounting for 83% of variation. Unfortunately all 60 variables have similar weightings in the 3 components. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function but 60 is too many, especially given the small sample. I

I wondered if, rather than use the 60 variables, itsit is possible to use the 3 principal components to build an LDF? I I am using STATAStata for the analysis. Does anyone have any suggestions? Thanks

I am interested in building a linear discriminant function using the best of 60 variables for a diagnostic test. I have calculated the area under the ROC curve for each of these individually and none have an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals. I have tried to reduce the variables using principal component analysis. There are 3 components accounting for 83% of variation. Unfortunately all 60 variables have similar weightings in the 3 components. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function but 60 is too many, especially given the small sample. I wondered if, rather than use the 60 variables, its possible to use the 3 principal components to build an LDF? I am using STATA for the analysis. Does anyone have any suggestions? Thanks

I am interested in building a linear discriminant function using the best of 60 variables for a diagnostic test. I have calculated the area under the ROC curve for each of these individually and none have an AUC greater than 0.73. I have a fairly small sample of 50 healthy and 50 diseased individuals.

I have tried to reduce the variables using principal component analysis. There are 3 components accounting for 83% of variation. Unfortunately all 60 variables have similar weightings in the 3 components. I would ordinarily pick the highest weighted variables and then incorporate them in a linear discriminant function but 60 is too many, especially given the small sample.

I wondered if, rather than use the 60 variables, it is possible to use the 3 principal components to build an LDF? I am using Stata for the analysis. Does anyone have any suggestions?

Source Link
Andrew
  • 51
  • 1
  • 2
Loading