Hello,

The odds ratio for a factor that contains more than two categories is interpreted as the ratio of the odds of the outcome for one category compared to the odds of the outcome for a reference category. The reference category is usually the one with the highest value or the most frequent value of the factor variable. For example, if race has three categories: white, black, and others, and white is the reference category, then the odds ratio for black is the odds of the outcome for black divided by the odds of the outcome for white and the odds ratio for others is the odds of the outcome for others divided by the odds of the outcome for white.

To interpret the odds ratios, you need to compare them to 1. An odds ratio greater than 1 means that the odds of the outcome are higher for that category than for the reference category. An odds ratio less than 1 means that the odds of the outcome are lower for that category than for the reference category. An odds ratio equal to 1 means that there is no difference in the odds of the outcome between that category and the reference category.

In your example, I guess that the race factor is the last one (jenisras) of which white is a reference category. Then, the odds ratios of the other two categories (hitam (black), lainnya (other)) are given against white. So, odds for black are 3.63 and for other are 2.64 times those of white and both are significantly greater than white with p < 0.05.

As Dr. Lyudmil mentioned,

you should determine one category as a reference based on your aim of study to compare other categories with it,

now you select jenisras as reference category and compare hitam & lainnya with it with an OR of 3.63 and 2.64 respectively, indicated that those last categories are at significant risk of low birth weight than the reference category,

But, when you set one of hitam & lainnyu as a reference category, the OR would be less than 1 and it's had a significant protective effect for lbw compared to those categories.

Lyudmil Antonov Hello, thanks for your answer! I would like to clarify? How do you set a reference group? By default stata set it up based on the lowest categorical number (i.e: 0= black, 1= white, 2=others, therefore the reference group would be black), however, theoretically, do we need to look up which group is more at risk for, let's say, lbw? or do we just do it based on the most frequent value like you stated before?

Issam S. Ismail Thank you. Now that you mentioned two scenarios which contain different racial group as the reference group, what is the most recommended way to choose a reference group? Thank you in advance!

There is no definitive rule for choosing a reference group, but some possible considerations are:

• Choose a level that is meaningful or relevant for your research question or hypothesis. For example, if you are interested in comparing different treatments to a control group, you may choose the control group as the reference level.
• Choose a level that has a large sample size or frequency, so that the estimates and standard errors are more stable and precise.
• Choose a level that is typical or representative of the population, so that the coefficients are easier to interpret and generalize.

In Stata, you can specify or change the reference level of a categorical variable using the i. prefix or the ib# notation in your regression command. For example, if you have a categorical variable x with levels 1, 2, 3, and 4, and you want to use 3 as the reference level, you can do:

reg y i.x // this will use 1 as the reference level by default reg y ib3.x // this will use 3 as the reference level

You can also use the fvset command to set the base level for a categorical variable permanently or temporarily. For example:

fvset base 3 x // this will set 3 as the base level for x permanently fvset temp base 4 x // this will set 4 as the base level for x temporarily

dear Fadhaa Aditya Kautsar Murti

it's based on your research objectives and what you find in literatures (previous studies).