How to test for an independent distribution?

[bengal007]

Hello All,

I would appreciate some advice on the following simple problem. I have a survey of internet users and want to test for the Null that the distributions of occupations of respondents across countries are identical.

To solve this, I have set proportions across countries and concluded that, in order for the distributions to be independent, ratios of occupations should all be identical. However, I'm quite sure my logic here is flawed. I have attached the table with the data and my workings in the .doc

Problem set.doc

Data table and my solution.

(43 KiB) Downloaded 540 times

and .xls file

Problem.xls

Further workings in Excel

(26.5 KiB) Downloaded 505 times

.

Thank you for your time.

[trubador]

I am not sure what you are trying to achieve, but I think a 2-way frequency tabulation analysis would be more appropriate for your data. Attached, you can find a workfile includes the series that I generated based on your problem set. The simplest way to carry out this analysis in EViews is to group the series first and then to select View/N-way tabulation...You can also use the following piece of code, if you like:

Code: Select all

valmap map1 valmap map2 map1.append 1 Canada map1.append 2 Europe map1.append 3 USA map2.append 1 Computer-related map2.append 2 Education/student map2.append 3 Management/prof country.map map1 occupation.map map2 group gr gr.add occupation country gr.freq(exp)

[bengal007]

Hello,

Thanks for your prompt answer. Grouping the variables and running a two-way frequency tabulation give me the result.
However, I was pretty certain I was a simple pen-and-paper approach to this. Please give me a shout if you can think of one.

Much obliged,
Dominik

[trubador]

It is indeed a very simple approach. See, for example: Wikipedia