Categorical covariates in parameter estimation

Question

Abed Alnaif 2023-11-3

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2042692-categorical-covariates-in-parameter-estimation

评论： Abed Alnaif 2023-11-24

采纳的回答： Arthur Goldsipe

Hello,

Do the SimBIology parameter estimation functions support categorical covariates?

Thank you,

Abed

2 个评论
显示无隐藏无

Joe Myint 2023-11-4

编辑：Joe Myint 2023-11-4

Hello Abed,

Take a look at this example and see if it is what you are looking for. This example shows how to estimate category-specific (such as young versus old, male versus female) parameters using PK profile data from multiple individuals.

https://www.mathworks.com/help/simbio/ug/estimate-category-specific-PK-parameters-in-a-hierarchical-model.html#EstimateCategorySpecificPKParametersMultipleIndividualsExample-12

Hope it helps, Joe

Abed Alnaif 2023-11-9

Thank you, Joe! I wasn't aware of this feature, and I think it will work for me. I'm going to try it.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Arthur Goldsipe 2023-11-6

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2042692-categorical-covariates-in-parameter-estimation#answer_1347422

Do you want to do nonlinear regression or nonlinear mixed effects (NLME)? The feature Joe mentions in his comment only applies to nonlinear regression. If you want to use categorical covariates with sbiofitmixed (for NLME), then there's no direct support for categorical covariates. I have an idea for a workaround, but I'd need to think about it a little more. If you are interested in the workaround, please let me know and I can investigate it a bit further. It would also be helpful to know more about your covariate model. How many categories are in your categorical covariate? What does your covariate model look like?

3 个评论
显示 1更早的评论隐藏 1更早的评论

Jeremy Huard 2023-11-23

编辑：Jeremy Huard 2023-11-23

在 MATLAB Online 中打开

Hi @Abed Alnaif,

I would recommend to convert categorical covariates into dummy variables or one-hot vectors, where each category is converted into a new binary column.

However, I would only use

categories and all zeros would encode for the last category.

Please note that I will be using a dataset shipped with MATLAB that does not contain time courses. But the categorical variable it contains will help illlustrate the general idea.

load patients
T = table(Age,Height,Weight,Smoker,...
          SelfAssessedHealthStatus,Location,...
          'RowNames',LastName);
T = convertvars(T,@iscellstr,"string")
T = 100×6 table
                Age    Height    Weight    Smoker    SelfAssessedHealthStatus             Location          
                ___    ______    ______    ______    ________________________    ___________________________

    Smith       38       71       176      true            "Excellent"           "County General Hospital"  
    Johnson     43       69       163      false           "Fair"                "VA Hospital"              
    Williams    38       64       131      false           "Good"                "St. Mary's Medical Center"
    Jones       40       67       133      false           "Fair"                "VA Hospital"              
    Brown       49       64       119      false           "Good"                "County General Hospital"  
    Davis       46       68       142      false           "Good"                "St. Mary's Medical Center"
    Miller      33       64       142      true            "Good"                "VA Hospital"              
    Wilson      40       68       180      false           "Good"                "VA Hospital"              
    Moore       28       68       183      false           "Excellent"           "St. Mary's Medical Center"
    Taylor      31       66       132      false           "Excellent"           "County General Hospital"  
    Anderson    45       68       128      false           "Excellent"           "County General Hospital"  
    Thomas      42       66       137      false           "Poor"                "St. Mary's Medical Center"
    Jackson     25       71       174      false           "Poor"                "VA Hospital"              
    White       39       72       202      true            "Excellent"           "VA Hospital"              
    Harris      36       65       129      false           "Good"                "St. Mary's Medical Center"
    Martin      48       71       181      true            "Good"                "VA Hospital"              

In the table above, the variable SelfAssessedHealthStatus contains 4 categories: Poor, Fair, Good, Excellent.

You can generate dummy variables for it with:

Tstatus = convertvars(T(:,"SelfAssessedHealthStatus"), "SelfAssessedHealthStatus","categorical");
Tstatus = onehotencode(Tstatus);
T = [Tstatus, T]
T = 100×10 table
                Excellent    Fair    Good    Poor    Age    Height    Weight    Smoker    SelfAssessedHealthStatus             Location          
                _________    ____    ____    ____    ___    ______    ______    ______    ________________________    ___________________________

    Smith           1         0       0       0      38       71       176      true            "Excellent"           "County General Hospital"  
    Johnson         0         1       0       0      43       69       163      false           "Fair"                "VA Hospital"              
    Williams        0         0       1       0      38       64       131      false           "Good"                "St. Mary's Medical Center"
    Jones           0         1       0       0      40       67       133      false           "Fair"                "VA Hospital"              
    Brown           0         0       1       0      49       64       119      false           "Good"                "County General Hospital"  
    Davis           0         0       1       0      46       68       142      false           "Good"                "St. Mary's Medical Center"
    Miller          0         0       1       0      33       64       142      true            "Good"                "VA Hospital"              
    Wilson          0         0       1       0      40       68       180      false           "Good"                "VA Hospital"              
    Moore           1         0       0       0      28       68       183      false           "Excellent"           "St. Mary's Medical Center"
    Taylor          1         0       0       0      31       66       132      false           "Excellent"           "County General Hospital"  
    Anderson        1         0       0       0      45       68       128      false           "Excellent"           "County General Hospital"  
    Thomas          0         0       0       1      42       66       137      false           "Poor"                "St. Mary's Medical Center"
    Jackson         0         0       0       1      25       71       174      false           "Poor"                "VA Hospital"              
    White           1         0       0       0      39       72       202      true            "Excellent"           "VA Hospital"              
    Harris          0         0       1       0      36       65       129      false           "Good"                "St. Mary's Medical Center"
    Martin          0         0       1       0      48       71       181      true            "Good"                "VA Hospital"              

Now, we can use Poor as the reference group, meaning that when all other groups are 0, then the category is Poor:

T.Poor(:) = 0
T = 100×10 table
                Excellent    Fair    Good    Poor    Age    Height    Weight    Smoker    SelfAssessedHealthStatus             Location          
                _________    ____    ____    ____    ___    ______    ______    ______    ________________________    ___________________________

    Smith           1         0       0       0      38       71       176      true            "Excellent"           "County General Hospital"  
    Johnson         0         1       0       0      43       69       163      false           "Fair"                "VA Hospital"              
    Williams        0         0       1       0      38       64       131      false           "Good"                "St. Mary's Medical Center"
    Jones           0         1       0       0      40       67       133      false           "Fair"                "VA Hospital"              
    Brown           0         0       1       0      49       64       119      false           "Good"                "County General Hospital"  
    Davis           0         0       1       0      46       68       142      false           "Good"                "St. Mary's Medical Center"
    Miller          0         0       1       0      33       64       142      true            "Good"                "VA Hospital"              
    Wilson          0         0       1       0      40       68       180      false           "Good"                "VA Hospital"              
    Moore           1         0       0       0      28       68       183      false           "Excellent"           "St. Mary's Medical Center"
    Taylor          1         0       0       0      31       66       132      false           "Excellent"           "County General Hospital"  
    Anderson        1         0       0       0      45       68       128      false           "Excellent"           "County General Hospital"  
    Thomas          0         0       0       0      42       66       137      false           "Poor"                "St. Mary's Medical Center"
    Jackson         0         0       0       0      25       71       174      false           "Poor"                "VA Hospital"              
    White           1         0       0       0      39       72       202      true            "Excellent"           "VA Hospital"              
    Harris          0         0       1       0      36       65       129      false           "Good"                "St. Mary's Medical Center"
    Martin          0         0       1       0      48       71       181      true            "Good"                "VA Hospital"              

During fitting, we can now use a covariate model such as:

Here,

(the population estimate) will correspond to the case where SelfAssessedHealthStatus = Poor,

will correspond to SelfAssessedHealthStatus = Fair,

will correspond to SelfAssessedHealthStatus = Good,

will correspond to SelfAssessedHealthStatus = Excellent.

I hope this helps.

Best regards,

Jérémy

Abed Alnaif 2023-11-24