Logistic regression with backward elimination in survey data

[Pawan Acharya]

Hi There,
I am using Demographic and Health Survey data and i want to perform logistic regression analysis (Dependent Variable; Institutional Delivery) with backward elimination approach. As the DHS is a two-stage cluster design, i need to adjust for sample weight and the primary sampling unit. I tried searching for the way of doing it but unfortunately, could not find a solution. The svy: command does not work for this.
Any guidance regarding performing backward elimination using survey data would be very helpful for me. Thank you.
Sincerely,

[Marcos Almeida]

Hello Pawan,

Welcome to the Stata Forum.

It seems we cannot perform backward regression under a survey design. Luckily for me, I am not a fan of backward regression...

Please read this FAQ:https://www.stata.com/support/faqs/s...-svy-commands/

Hopefully that helps.

Best,

Marcos

[Carlo Lazzaro]

Pawan:
as you may find out yourself in different threads on this forum,the warmest recommendation about backward elimination (and stepwise approach in general) is: don't do it!