Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - georgeha

Pages: [1]
Good suggestion. After running a second step with the first step's anchored parameters, I see the sequence is correct (the difficulty estimates match the original item numbers). So, the Delete!Item(n) does not appear to be messing up the sequence.  This is good.

I'm examining a within-item multidimensional model, so I am exporting the item parameter file to specify these as initial values in the next step.  I am excluding the 2nd item, however, because it is not part of the model I am investigating (but I wish to retain the item numbering I have).  If I exclude this item, using "DELETE!items(2)" does the second step read a disordered set of initial parameters (offset due to the deleted item)? The error, "If anchors or initial values are used the referencing must reflect removal of this parameter" to me suggests the ordering gets messed up. 

If the initial values are indeed disordered, what is the easiest way to ensure the ordering is correct?

Questions and Answers / Re: Fit indices in ConQuest
« on: December 02, 2013, 12:57:07 AM »
Hi Navid,

I'll give my best shot at answering your questions and see if anyone negates what I say.

1) Yes, the deviance statistic is the -2LL estimate (there's no need to take its log or multiply it by -2).
2) AIC = deviance + 2(n_parameters). BIC = deviance + Ln(N_persons)(n_parameters).
3) Good question.  Does anyone know?  It would make sense to me to consider the testlet model as nested since you're not modeling within-item multidimensionality. However, I'm not sure about comparing deviance statistics when you model inter-item dependence. I hope someone answers this.


Questions and Answers / Re: Including Case PID in Outputs
« on: December 01, 2013, 08:22:21 PM »
Ah, thanks, Eveline. Good to know. I hope that after I purchase the software (I'm using the trial until my uni approves purchase), I can get updates when they come out. You're right--it doesn't really crash, it just keeps running.  I tried letting mine run for an hour (on my 16GB 2.9Ghz machine), but it still didn't finish, so I stopped it.

The other reason I wanted to keep PIDs was that I am comparing this unidimensional model with multidimensional models. In my multidim model, some of the cases are missing data on all the items in one or more of the dimensions.  For making model comparison, these cases should be eliminated from both the multidim and unidim models before I get my -2LL estimates.  Having a stable PID will lower chances of mistakes, but now that I know this is a bug, I can work with the ordering and just be extra careful :-)



Questions and Answers / Including Case PID in Outputs
« on: November 27, 2013, 08:43:18 PM »
My objective is to keep the ID from the data file connected to the person estimates in the outputs from the Show Cases (and from the Export Scoreddata) commands (so I can subsequently read the output into SAS easily, thereby connecting the estimates to my other datasets from other analyses of these persons). To do this, I assumed, after reading the manual, I need to specify PID in the FORMAT command, but when I specify PID (as FORMAT PID 1-8 time 9 responses 11-40), ConQuest crashes. If I specify it with any other name, such as "ID" (as in this: (as FORMAT ID 1-8 time 9 responses 11-40), ConQuest runs fine, but the output ID is just the order ID, not the 8 digit ID from my data file.
I read that ConQuest has to sort the cases when PID is used (they are already sorted), but I only am examining about 500 cases, so I wouldn't expect a sort to crash the program.  The model is unidimensional and converges fairly quickly (79 iterations in less than 10 seconds).

Am I missing something?


Pages: [1]