HELP! how to test proportions and means in 2 cross sections?

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

gaalcaraz
Posts: 4
Joined: Mon Oct 20, 2014 7:07 pm

HELP! how to test proportions and means in 2 cross sections?

Postby gaalcaraz » Mon Oct 20, 2014 7:12 pm

Hello!
I want to test whether some proportions and means are statistically different (if that is the right test) or not, between 2 cross sectional survey times (2006 and 2011).
For instance, i want to test whether school enrollment is different or not in those 2 years, and the average age on which school drop out occurs.
Can any one suggest me the correct test for this purpose?
many thanks!
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: HELP! how to test proportions and means in 2 cross secti

Postby GerineL » Tue Oct 21, 2014 8:26 am

assuming you have the same cases in the 2 years, look into t-test for paired samples. If they are 2 independent samples, look into t-test for independent samples.
If this does not help you, please provide more info on your data.
gaalcaraz
Posts: 4
Joined: Mon Oct 20, 2014 7:07 pm

Re: HELP! how to test proportions and means in 2 cross secti

Postby gaalcaraz » Tue Oct 21, 2014 12:55 pm

Hello!
Thanks!

No, they are totally unrelated samples.
they are nationally representative household surveys, which were carried out in 2006 and in 2011.
I want to compare a couple of indicators from both surveys, in order to see if there was any change from one survey year to the other one.

so, t-test independent samples would work for the averages/means..
what would work for the proportions?

would I need a pooled dataset in order to make the test? (I mean, joining both surveys in one single data file)
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: HELP! how to test proportions and means in 2 cross secti

Postby GerineL » Tue Oct 21, 2014 1:36 pm

gaalcaraz wrote: so, t-test independent samples would work for the averages/means..
what would work for the proportions?
could you give an example of one of the proportion outcomes?
would I need a pooled dataset in order to make the test? (I mean, joining both surveys in one single data file)
Yes, you would. I assume you have the same variables in both datasets (otherwise you could not compare them of course).
So you could add a variable "year" to both datasets.
You wuold give this the value "2011" for 2011, and "2006" for 2006.
Using the merge files - add cases command, you can add the 2 files together.
gaalcaraz
Posts: 4
Joined: Mon Oct 20, 2014 7:07 pm

Re: HELP! how to test proportions and means in 2 cross secti

Postby gaalcaraz » Tue Oct 21, 2014 2:17 pm

Thanks!

The proportions are for instance:
-proportion of unemployed persons in the age range 18-30
-proportion of children (7-17 years) who work
-proportion of households receiving government assistance
-proportion of persons working in the agricultural sector
-proportion of persons (over 18 and currently working) concerned of becoming unemployed in the next 12 months
-proportion of poor households
-proportion of households reporting having insufficient income for covering their basic needs
-proportion/share of food expenditures, out of total household expenditures
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: HELP! how to test proportions and means in 2 cross secti

Postby GerineL » Tue Oct 21, 2014 3:03 pm

it depends a bit on the level of your variables (e.g., there is a difference between nominal variables (e.g. yes/no variables like do you receive support yes or no?) (or: gender) (or: religion) or ordinal level variables (e.g., education level))
most of your questions seem to be nominal (yes/no), so you could use crosstabs with chi square for that.
gaalcaraz
Posts: 4
Joined: Mon Oct 20, 2014 7:07 pm

Re: HELP! how to test proportions and means in 2 cross secti

Postby gaalcaraz » Tue Oct 21, 2014 4:26 pm

a very very very BIG THANK YOU!!! :D

Who is online

Users browsing this forum: No registered users and 3 guests

cron