Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Mokken scaling procedure - output

From   Nick Cox <>
Subject   Re: st: Mokken scaling procedure - output
Date   Tue, 12 Feb 2013 16:53:02 +0000

This is way outside my experience, but I would not exclude plotting
the data, seemingly unusual in this territory. In this case, -graph
matrix- might be helpful but given the ordinal scale jittering would
be essential.

set obs 1000
forval j = 1/5 {
      gen y`j' = floor(6 * runiform())
raph matrix y?, jitter(1)

gives a kind of benchmark for a null extreme.


On Tue, Feb 12, 2013 at 4:44 PM, Laura Maria Schwirz <> wrote:
> I thought so too. I was quite surprised to get such high Loevinger
> coefficients but this might be due to a rather large sample size of 3,300?
> Also, does it mean that if they all lie above the critical value of 0.3
> that these variables (or traits in this case) form one dimension? I.e. if
> two of them were lower than 0.3, the other two would still form such a
> dimension?
> Many thanks for your advice on this, much appreciated!
> Laura
> On 12 February 2013 16:26, JVerkuilen (Gmail) <> wrote:
>> On Tue, Feb 12, 2013 at 10:42 AM, Laura Maria Schwirz <> wrote:
>>> Apologies. I had some trouble sending the email and I thought the
>>> Stata output caused the problem. Here's the output. I am actually just
>>> reading van Schur's article and find it quite helpful.
>> Your situation is exactly what Mokken scaling is good at but read
>> through those citations to make sure you feel solid about what the
>> output is.
>> I'd worry about a halo effect on these due to the likely content.
>> Those are some VERY high H stats.
*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index