+ Reply to Thread
Results 1 to 4 of 4

Thread: Statictical suggestion regarding DOE and GLM for the pharmacutical related problem

  1. #1
    Points: 69, Level: 1
    Level completed: 38%, Points required for next Level: 31

    Location
    UK
    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Cool Statictical suggestion regarding DOE and GLM for the pharmacutical related problem




    We performed a DOE to study the relationship between simulated intestinal fluid (SIF) components and solubility of drugs. The DOE is composed of 7 factors (representing the components of SIF) and 2 levels for each drug. And we have done several drugs with the DOE. Now with the DOE analysis, we can tease out significant factors for each drug, and now we want to compare all those drugs by correlating the significant factors with the drug properties (multivariate).



    At the beginning, we want to see how each factor can affect the solubility. Actually the increase/decrease of solubility if the factor increase (therefore difference of solubility between high/low level). So we use DoE and it can tell us the significance of each factor within the same drug. Most of the application of DOE stops here as they only concern one response (like productivity) related with cost or time, but now we have a series of drugs, and we want to get information not only within the same drug, but also between drugs. For example, pH is a significant factor for drug A, but not drug B and we want to dig out the reason why, like different propertyies of the drugs, like pKa. So that is where we start to have problems.

    The drug properties (pKa, log P, molecular weight ......) are sort of independent variates, yet some of them may have partially overlapping information. Currently we use DOE analysis for individual drugs and GLM to pool all the information together. For GLM we use solubility as response, 7 factors and adding the drug properties as covariate to scale up a model. But we still find this is not the best way to do the statistics, as larger valus of certain drug may perdominate the analysis and we lost information of outliers.

    So I just want to ask does anybody come up with other possible analysis method for the multivariates correlations??

    Any thoughts or suggestions or questions are welcome

  2. #2
    Point Mass at Zero
    Points: 13,828, Level: 76
    Level completed: 45%, Points required for next Level: 222
    ledzep's Avatar
    Location
    Berks,UK
    Posts
    684
    Thanks
    188
    Thanked 143 Times in 139 Posts

    Re: Statictical suggestion regarding DOE and GLM for the pharmacutical related proble

    "But we still find this is not the best way to do the statistics, as larger valus of certain drug may perdominate the analysis and we lost information of outliers."

    Why not log-transform the response? Does it not help? Is it still flagged as an an outlier (on residuals plot) when analysed on log-scale?
    Oh Thou Perelman! Poincare's was for you and Riemann's is for me.

  3. #3
    Points: 69, Level: 1
    Level completed: 38%, Points required for next Level: 31

    Location
    UK
    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Statictical suggestion regarding DOE and GLM for the pharmacutical related proble

    Name:  1.jpg
Views: 20
Size:  25.5 KB

    Thank you for your reply!
    Still have a list of outliers, as you can imagine the solubility value (response) is really random. And I am not sure whether it is sensible to do the log-transform.

  4. #4
    TS Contributor
    Points: 14,811, Level: 78
    Level completed: 91%, Points required for next Level: 39
    Miner's Avatar
    Location
    Greater Milwaukee area
    Posts
    1,171
    Thanks
    34
    Thanked 405 Times in 363 Posts

    Re: Statictical suggestion regarding DOE and GLM for the pharmacutical related proble


    What happened around observation 290? Looks like you may have a lurking variable or covariate.

+ Reply to Thread

           




Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats