Multivariate logistic Regression with binary predictors

I have a dataset of 10 independent variables describing whether a symptom of a disease is present or not and then the binary dependent variable of whether the specific disease is present or not. There are approximately 1000 samples, however some of them are missing values for some of the independent variables.

Is it possible to perform multivariate logistic regression with this dataset to predict disease presence or absence? If so, how would I go about doing that in MATLAB?

Thanks in advance!

Answer 1

[Shashank Gupta]

Hi,

In my opinion you can apply multivariant logistic regression, but you need to take care of the blank or missing values in between, In order to train the model, either you can ignore that particular sample if it has missing value or you can ignore the specific dimension. Also, If you have any prior knowledge of what could be the missing value, you can fill up the data by interpolating it,

There is function “mnrfit”, it’s a multinomial logistic regression function, May be this function can give you a head start where to look for.

I hope it helps.

Answer 2

[Mateus Bringel Oliveira Duarte]

Hi,

In General Matlab has some difficult to handle with some typical problems in medical research.

Logistic regression is considered a geneleralized linear model, so you can try the function fitglm (https://www.mathworks.com/help/stats/fitglm.html). Note that to extract the OR you can Exponentiate the Variables Estimates ,EXP("x"), It's is crucial you also report the Confidence intervals, and P-value.

Best,