Help with statistic model
1
0
Entering edit mode
16 months ago
Bioman • 0

I have a bunch of statistic question that are very smiler to this question

"Based on the model, what is that probability that a 36 old male individual with a BMI of 24.5 is diagnosed with T2D using a OGTT?"

or question where you are asked to get an idea on a persons predicted BMI based on other factors.

I can't find anywhere online that can help me understand how to solve these forms of questions

Im not asking anyone here to come with answers, just buzzwords I could google that can teach me how to solve such question

I know you need the odds ratio, and I can calculate it using R and general logistic model (glm()) but i can't figure out how to use that information for further data analysis.

online education R • 679 views
0
Entering edit mode

I think if you have odds ratios - you are golden. Odds ratios is the ratio between the chances to get T2D and the chances not to get T2D. Together, events like "T2D" and "not T2D" comprise all the probabilistic space, so if you have OR of 1/4 - the probability will be 0.2 (since 0.2 / (1 - 0.2) = 1/4 - simple equation with one unknown x).

0
Entering edit mode
16 months ago

OGTT is the oral glucose tolerance test, if I recall.

I am guessing that you first have to construct a model and then make predictions from this. For the question posed, I suppose that the model would be:

m <- glm(
T2D ~ OGTT + BMI + Age,
data = mydata,
family = binomial(link = 'logit'))


Then, make predictions using this model by setting values for BMI (e.g. 24,5) and Age (e.g. 36) - you may have to vary the values of OGTT while doing this, i.e., you will produce different probabilities based on different values of OGTT, but fixed values for BMI (24,5) and Age (36). Predictions are made via:

predict(m, type = 'response', newdata = ...)


I think that we go over this exact topic in our Lecture 3 notes: https://github.com/kevinblighe/Rtutorials Sorry, these are still in PowerPointÂ® and not updated to the more fancy / exquisite R Bookdown / Markdown.

Kevin