www.cmm.bristol.ac.uk/forum

HI there,
I was just wondering if it is possible to get the estimates of the fixed effects parameters in Odds ratio or Relative risk when using logistic or Poisson regression?
Many thanks
Bea

Good question!

Answer is no not currently, but this should be easy to add. We will look
into implementing this asap.

Hi Bea,

We have now implemented options to display the fixed part parameter estimates in many of runmlwin's multilevel discrete response models in their exponentiated form:

(1) Use the or option to get odds ratios in multilevel binomial response logit models and in multilevel ordered response multinomial models.

(2) Use the irr option to get incidence-rate ratios in multilevel count response poisson or negative binomial models.

(3) Use the rrr option to get relative-rate ratios in multilevel unordered response multinomial models.

Standard errors and confidence intervals are similarly transformed.

To get these new options, you will need to download the latest version of runmlwin. To do this, simply type the following in a net aware version of Stata:

. adoupdate runmlwin, update

The first thing you should do is read the new entries for these options in the help file:

. help runmlwin

Let us know if you have any problems.

Best wishes

George

Hi,

I am using the OR option with a multilevel binomial response logit model, but I am having some problems with the Confidence Intervals. I get the following results:

Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
1.629818 .3478006 2.29 0.022 .9475621 2.312074

The estimate and SE for that variable in MLwiN are 0.448(0.213), but according to the CI, this is not statistically significant?

The same has happened in other models:

Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
1.637491 .343867 2.35 0.019 .9629507 2.31203

Could you please advice?

Thank you,
Laia

Hi Laia,

Thank you. You spotted a bug. We were not calculating the lower and upper bounds of the 95% confidence intervals for odds ratios in the correct way. We were calculating them as plus or minus 1.96 times the standard error of the odds ratio. This is incorrect as it assumes the sampling distribution of an odds ratio is normal.

The correct way to calculate the lower and upper bound for the 95% confidence interval of an odds ratio is to simply calculate the exponential of the lower and and upper bounds of the associated regression coefficient.

We have now fixed this.

To get the latest version of runmlwin, simply type the following in a net-aware version of Stata:

. adoupdate runmlwin, update

Best wishes

George