Messages

1

Questions and Answers / Re: Estimating the Rasch Testlet Model

« on: December 16, 2012, 11:34:16 PM »

Yes Davide, Steffen's papers are quite good and detailed explanations of how to accomplish this.  Perhaps you could email him directly for more detailed assistance.
Ray

2

Questions and Answers / Re: ConQuest freezes when trying complex between-item multidimensional model

« on: October 23, 2012, 01:26:37 AM »

You are using an 8-dimensional model with default estimation.  That means 15 to the power 8 nodes.  My apologies that it is not reporting a more informative error message, I'll fix that.

What you should be doing is using montecarlo estimation, not gauss-hermite quadrature.

3

Questions and Answers / Re: R square in latent regregression

« on: October 07, 2012, 08:56:53 PM »

luis, the formula is in the previous post

4

Questions and Answers / Re: Residuals from the itemss

« on: October 07, 2012, 08:55:54 PM »

residuals are defined at the level of a combination of student and item.  ie the residual for item x and person n.  These are available from the show command.  For a summary of residuals by item you have the item fit statistics when you request  show items.

Ray
 

5

Questions and Answers / Re: Re-reunning new analysis with initial values

« on: September 24, 2012, 11:55:48 PM »

sounds to me like the results are stable and you can use them.  Have you exported historyu file and plotted trajectories for the parameter estimates and deviances?  I find that helps me with confidence in the parameter estimates.

6

Questions and Answers / Re: R square in latent regregression

« on: September 24, 2012, 11:52:53 PM »

Luis, to get the R-squared do a run with regressors and one without.  In each case you'll get an estimate of the (conditional) variance of the latent variable.  Let's call V, the variance without regressors and VR the variance with regressors.  The R-squared = 1-VR/V=(V-VR)/V

ie the proporation of total variance that is explained by the regressors.

7

Questions and Answers / Re: Error: too many levels

« on: September 23, 2012, 05:56:05 AM »

are you using a demo version?

8

Questions and Answers / Re: Re-reunning new analysis with initial values

« on: September 23, 2012, 05:55:16 AM »

It means the parameters may not be well estimated.  Try again with more nodes.  Yes the import command is used to start of the estimation with your current estimates.

9

Questions and Answers / Re: EAP file

« on: September 23, 2012, 05:53:46 AM »

first field is EAP estimate, second field is the posterior standard deviation and third is reliability of the estimate.

If you want a point estimate EAP is often my prefered estimate.

10

Questions and Answers / Re: EAP's and PV's

« on: July 24, 2012, 01:41:18 AM »

These are huge questions.  I will do a bit of work on a note that explains the differences, but it will take a while.  In the mean time doing some reading on the web and from other available resources might help.

here are a few points.

1.  maximum likelihood (MLE) and weighted maximum likelihood (WLE) are point estimates that depend only upon the item response model.  The WLE is Warm's improvement on the MLE that is close to an unbiased estimator.

2.  EAP is expected a-posterior, or the expectation of the posterior.  I believe the term was introduced by Bock.  The EAP is a minimum mean squared estimator that is in general biased.  The EAP depends upon both the item response model and the population model. 

3.  PV's are plauible value, they are random draws from the posterior and should be used whenever you wish to undertake analyses with the ability estimates (ie estimate population parameters -- means, variances, covariances, regressions, t-tests and so on)

Ray

11

Questions and Answers / Re: Saving Mantel Haenszel output

« on: July 24, 2012, 01:29:40 AM »

David,

note that you can use

set buffersize=a big number;

to make sure you don't lose material from the output window.  Xiaoxun is looking into the export issue.

Ray

12

Questions and Answers / Re: Displacement available in ConQuest?

« on: July 24, 2012, 01:28:06 AM »

Sukkeun,

ConQuest does not provide this value directly, but here is a way to get it.

1. Do the estimation with the items anchored and save an item estimates file.
2. do a second run using the item estimates file as anchors, but omit the items that were previously anchored.  request just one interation in the estimate command.
3. The output from the second run will show how much shift in the originally anchored items would have been

if you'd like me to assist in setting up such a run please email me your files  conquest@acer.edu.au

Ray

13

Questions and Answers / Re: Getting error

« on: July 10, 2012, 10:42:05 PM »

Hi Cathy, is this now sorted following our email exchange?

14

Questions and Answers / Re: When constarin correlations to zero?

« on: July 04, 2012, 02:24:50 AM »

In a between item context constraining them to zero is just like doing separate runs, so I can't think of many contexts in which you'd want to do it.

15

Questions and Answers / Re: zero correlation

« on: July 04, 2012, 02:22:44 AM »

Allowing a correlation is a more general model, so it should fit better.  Fixing the correlations at zero has the same effect as running each of the scales separately.  If they are correlated you'll increase reliabilitly because you'll be drawing strength across dimensions.