Missing items and discrimination

[Dany]

Hello, I have tried a lot, but I didn't succeed in getting the missing items and the discrimination under conquest. Can anyone help me please?

[dan_c]

I'm not sure what you mean by "getting" the missing items. Item discrimination is reported in the output from a call to itanal.

Can you provide an example of what you are trying to do?

See this post for more info:

Some users want to ask questions, report bugs, or provide examples of functionality in ConQuest.

To make this easy and to ensure the members of the forum can follow along and reproduce the issue (and maybe contribute their own solutions), it would be helpful to post a minimum working example.
...

[Dany]

Hello,
Thank you for your reply.
what i mean is the missing item. the item that has been missed in the test, Attached is an example from winstep, and i am trying to get this information in conquest, can I? (MISFITTING PERSONS OR ITEMS and OVER FITTING ITEMS).
Regarding the discrimination, it is not shown in my itanal output file. attached are my .cqc file and itanal file.
I have uploaded the one with wrong model statement, i am using (model item; ).

I have an error also and I cannot manage it. attach is a screenshot of the error. (Parameter(2) for item 2, is not identified)
How to solve this error?
Thank you.

[dan_c]

Thanks for attaching the example.

I think there are a few small errors in your uploaded file, but if you fix these you will get all the information that you need.

In your cqc file, you have the following model statement:

model item - dif.variable + item*dif.variable;

You have not defined the variable dif.variable anywhere and so ConQuest stops. Check out page 85 of the Command Reference for information about how to specify your model statement https://www.acer.org/files/Command-Reference.pdf. Can you provide some more information about the kind of model you are trying to estimate?

To your questions:

• missing

Your call to show will report the number of cases included in the analysis. You'll find the proportion of missing for each item in the output from itanal. In your example, because you used the option format=summary, the column N reports the number of valid responses for each item.

• discrimination

Your call to itanal reports the discrimination. See the column Item-Rst Cor which is the corrected item point-biserial (the correlation between the item score and the total scores by all of the rest items in the test (exclusive the item)) read more here https://www.acer.org/files/Conquest-Notes-5-ItemPointBiserialDiscrimination.pdf

[/list]

[Dany]

Hello, thank you for replying,
I have qcm questions which are divided into two levels; low and intermediate. My model statement is (model item;)
I have 21 students and 32 items. After I made my data file composed from 21 row of data (students' id and their responses), and my label file composed from 32 items, in the estimation I got a warning (parameter (2) for item (2) is not identified) and for more items the same warning too. My cqc file is the same as the one i uploaded before, but I have changed the format statement and the model one ( model item;).
I uploaded a screenshot for the warning in the question before.
Please, I would like to know where is the problem so that I have this warning and how can I solve?
Actually I have spent one week learning conquest, getting and interpreting the data I need, but with my real data (21 students and 32 items) I am getting this warning, and it leads to removing the items from estimation.
Thank you for telling me how can I solve it.
Highly appreciated. 

[dan_c]

I think your error probably relates to your data - the working example you posted works with no errors with Example 7 data and specification.

Can you upload your data? If not, can you send it via email. You can find my email address in my profile.

[Dany]

Hello,
Thank you for replying. Attached are my data for your review please.
I cannot manage the problem that I am facing.
Thank you so much for reviewing. 
highly appreciated.

[dan_c]

I don't get an identification error running this model. What version of ConQuest are you running?

I do get an error, Covariance matrix is not positive definite This often indicates problems in the data or model.

If you remove the dimensionality and estimate a 1 factor model, your item MNSQ are highly variable and some are quite poor. See e.g., https://www.rasch.org/rmt/rmt83b.htm

Other problems could be:

• a negative variance or residual variance for a factor (2 in this case)
• a correlation >= 1 between your factors
• or linear dependencies amongst your items (e.g., items that are perfectly correlated) 

I think you need to consider if you are specifying the right model given the data (have you conducted an EFA?), and is you have sufficient data given the responses.

[Dany]

Hello,
No, I didn't use EFA.
I have version 4.5, I have downloaded it from the website: https://www.acer.org/conquest. I have that warning each time I estimate the data.
What can I do for these data?
Actually, these are my real data I have and the answers from students.
How can I get the results of these data?
Is there any solution please?
Thank you in advance.

[Alvin Vista]

I think you have insufficient data. You're attempting to fit a 2-dimensional model on just 21 cases.