Hi,

In your study/research you are using Binary Logistic Regression only.  Take dependent variables and independent variable.  And do the analysis in SPSS software.  For that in google so many references and in videos are available in youtube also.  Kindly see who to do the analysis.  After finding the variables in the equation, its Beta values, SE, Significant values, and 95% confidence intervals respectively.  Then, you have to take all the beta coefficient values and the x - values as per your analysis.  Then you have to subtitute all the values in the Binary Logistic Regression equation, then you will get the odds ratio value.  Then you can find the probability value of a particular disease or etc.

Kindly see the following papers:

https://www.researchgate.net/publication/232767415_Prediction_of_Iodine_Deficiency_Disorder_in_school_going_children_in_the_coastal_areas_of_Puducherry__using_Binary_Logistic_Regression_approach?ev=prf_pub

https://www.researchgate.net/publication/216427254_Prediction_of_diabetic_retinopathy_among_diabetics_using_binary_logistic_regression_approach?ev=prf_pub

Article Prediction of Iodine Deficiency Disorder in school going chi...

Article Prediction of diabetic retinopathy among diabetics using bin...

Thanks for prompt response. I was wondering if I will have to include the result of the variable am controlling for in the result table, since it was not in the initial model. I will happy if you can clarify. Thanks

you can report it like in two models and it is good for the comparison purposes.like the first model (model1) with the unadjusted model (full model) and second model (model2) the one you control for confounding variables.

Here is a reference on how to present the results from a multivariate binary logistic regression: http://www.indiana.edu/~jopeng51/teaching-logistic.pdf

What you talked is model fitting or diagnosis. The basic answer is yes. The Key issues in model fitting are:

1.       A model should be complex enough to fit well

2.       but also relatively simple to interpret, smoothing rather than over fitting the data.

The potentially useful models are usually a small subset of the possible models, which includes a certain terms: A study designed to answer certain questions, such as treatment group;  and the models should recognize distinctions between response and explanatory variables. Hence

1.       The modeling process should concentrate on terms linking responses and terms linking explanatory variables to responses.

2.       The model should contain the most general interaction term relating the explanatory variables.

If a factor showing significance and it is uncorrelated with other factors, it should be included. The deviance (-2*log likelihood) difference (G2) between the model with and without such factor will be significant.

If that factor is correlated with other factor, especially with main predictor, we need consider the co-linearity, adding such factor, the estimation for the main predictor is biases, and the standard error for the model might be inflated.

In addition, we also need consider the interaction between that factor and the main predictor. If the interaction exists, (for interaction, 10% significant level might be considered to be included), definitely both that factor and its interaction should be in the model.

Thank you for explanation.

Akinboro