[muhammad waleed]

Dear all Not answering to above asked question just posting a similar question,

i am running an analysis to check association between LBW (low birth weight, binary variable for child) and poverty for various states. LBW is dependent variable. in independent variables I have poverty status of household, mother education, sex of child, area of residence of household (HH), income of household.

in my case i Have multiple childs information for same household. So i think in my data children are nested in HH because some HH specific characteristics remains fixed for all children belonging to same HH or who are siblings. For that matter I am using following command to run melogit.

melogit LBW i.(sex educ_mother poor area) income || hid:

I am running this model separately for five states. But I am confused now about output and have some questions for you.
like in logit we interpret marginal effect and not coefficients mostly, what will be interpreted in melogit?

1. how to get marginal effects in melogit
2. how to standardize coefficients for melogit in order to compare magnitudes of variation from different sources. (I have to compare results for five states in end)

Its urgent, I will really appreciate an early response.

[Clyde Schechter]

like in logit we interpret marginal effect and not coefficients mostly, what will be interpreted in melogit?

This is not true as a generalization. Actually, it is fairly controversial. While I personally prefer, in most circumstances, to look at marginal effects over coefficients or odds ratios, many people feel strongly to the contrary. Either choice will find supporters and detractors.

how to get marginal effects in melogit

Use the -margins- command with the -dydx)- option. But bear in mind that because the logistic model is non-linear, the marginal effect of a variable depends on the values of all of the variable in the model (including the variable whose marginal effect you are calculating). So you need to decide what reference values of variables to use for the marginal effects, or decide to use average marginal effects. Make these choices carefully as the results may have very different interpretations.

ow to standardize coefficients for melogit in order to compare magnitudes of variation from different sources

Please don't do that. It's a terrible idea. Given that most of your predictors are dichotomous standardizing the variables is even worse than it normally is. Moreover, the notion that standardization enables you to compare the contributions of different sources of variation is a fallacy anyhow.

I have to compare results for five states in end

The simpler way to do this is to include all five states in a single model and interact the state indicators with the other variables.

melogit LBW (i.(sex educ_mother poor area) c.income)##i.state || hid: