Dr. Ameer, Thank you very much for making this awesome explanation of how to use MFRM. You've given me confidence to use this tool in a project that I'm working on. The best part of this tutorial is the many examples you give and the way you explain relatively difficult concepts in very understandable terms. I recommend to anyone who wants to learn about MFRM to watch this tutorial.
This is so helpful, and thank you so much for sharing!!! It literally worths a semester of Rasch measurement classes~~~~Talking about FREE education!!!
I really enjoyed the presentation on Multi-Facet Rasch analysis. What an eye-opener! I wanted to see the datafile more closely but the font size is too small! Possible to access a more legible datafile? Will appreciate, otherwise I found the exposure extremely helpful!!!
Hi Saleh, thanks for the very nice lecture on Rasch. thumbs up! I got a bit lost at the wright map part, esp. at 25:10. if the students are one measurement below, why the possibility of answering the question is 0.27?
Hi. Think of any test as an interaction between the items on the test (specifically their difficulty) and test-takers (specifically their ability on whatever the test is measuring). If a test-taker has a very high ability, and comes across an item that is very easy (low in difficulty) then there is a very high probability that the test-taker answers this item correctly. Conversely, if a test-taker with very low ability comes across an item that is very difficult, the probability of answering correctly is very low. Now think of the Wright map as a ruler, but rather than measuring the distance between point A and point B in centimeters/inches, it measures distance in logits. In addition, rather than measuring only one thing, it is measuring two things at the same time (student ability and item difficulty). With the example used in the video the logits goes from -4 (at the bottom) all the way to +4 (at the top). There is a vertical line in the middle. The test-takers are on the left side of that line and are arranged according to their ability. Test-takers at the bottom, at -4 logits on the map (identified by the #) were technically the ones with the lowest ability. And at the top, at +4 logits there are 6 test-takers (remember, each # means 2 test-takers, and at the very top were 3 #'s), these were the students with the highest ability. On the right side of the vertical line are items, also arranged according to their difficulty. Item R04 at the bottom, at -3 logits was the easiest, whereas item R17 was the highest item on the map , and hence most difficult. This distance (in logits) between a test-taker and an item tells us the probability of that test-taker answering that item correctly. So, if the item (in terms of difficulty) is at the same logit of test-taker ability, then there is a 50/50 (50%) probability of answering correctly. If the test-taker measures one logit below the item, so for example if the item is at 0 logits and the test-taker is at -1 logits then, mathematically, the probability of answering that item correctly is .27, or 27%. How this number is calculated is covered at minute 27:00 in the video. Clear as mud? :)
Dr Saleh, this is very helpful, would like to ask a question regarding R setting for the maximum score, in the video is stated 9, video 2:39:41 it is assuming that all the score for the questions is the same maximum score? Can it be done for a different range of marks for different question? Say question 1 is score 1-4, question 2, 1 to 6 do i put R as 6?
Hi Ignatius, awfully sorry for the late reply. I would suggest you do a partial credit model where each item has its own rating scale defined. So your specification file would look like this: Models= ?,?,#,R6 We specified R6 because that was the highest possible score in your example. I discuss partial credit models in the last part of the video at 03:00:48 Let me know if this was useful.
Thanks for this. Very long, but very useful. I have a q: how is a skill like writing unidimensional? Surely it is so complex that it is multidimensional!
You're welcome. Good Q. We need to distinguish between psychological unidimensionality/multidimensionality and psychometric (or measurement) unidimensionality. You are right about writing (or any language ability for that matter) being multidimensional in terms of psychology (e.g., many psychological processes are involved when we write), but in pure measurement terms if these multiple psychological processes are used in unison, the assumption of unidimensionality will hold. See McNamara (1996) and Barkaoui (2014) (both in the reference list on the video) for more.They both discuss this specific issue in greater detail.
OMG !!! my favorite professor ever !! LOVE YOU SO MUCH DR.Saleh 🤍💕💕💕💕💕💕💕💕💕
Thank you for your kind and detailed lecture. It is the best online lecture about Rasch Model we can find on RUclips.
You’re most welcome. I’m glad you found it useful
Dr. Ameer, Thank you very much for making this awesome explanation of how to use MFRM. You've given me confidence to use this tool in a project that I'm working on. The best part of this tutorial is the many examples you give and the way you explain relatively difficult concepts in very understandable terms. I recommend to anyone who wants to learn about MFRM to watch this tutorial.
Appreciate your kind words.
Very educational! Thank you so much! ❤
This is so helpful, and thank you so much for sharing!!! It literally worths a semester of Rasch measurement classes~~~~Talking about FREE education!!!
Thank you for your kind words. Glad you found it useful :)
Love your teaching method
Thanks a lot, Professor. I learnt a lot.
My pleasure :)
great content. helped me a lot with my thesis.
I really enjoyed the presentation on Multi-Facet Rasch analysis. What an eye-opener! I wanted to see the datafile more closely but the font size is too small! Possible to access a more legible datafile? Will appreciate, otherwise I found the exposure extremely helpful!!!
Qetelo Moloi Thank you. Drop me an email and I’ll send you the data file. dr.saleh.ameer@gmail.com
Hi Saleh, thanks for the very nice lecture on Rasch. thumbs up!
I got a bit lost at the wright map part, esp. at 25:10. if the students are one measurement below, why the possibility of answering the question is 0.27?
Hi.
Think of any test as an interaction between the items on the test (specifically their difficulty) and test-takers (specifically their ability on whatever the test is measuring).
If a test-taker has a very high ability, and comes across an item that is very easy (low in difficulty) then there is a very high probability that the test-taker answers this item correctly. Conversely, if a test-taker with very low ability comes across an item that is very difficult, the probability of answering correctly is very low.
Now think of the Wright map as a ruler, but rather than measuring the distance between point A and point B in centimeters/inches, it measures distance in logits. In addition, rather than measuring only one thing, it is measuring two things at the same time (student ability and item difficulty).
With the example used in the video the logits goes from -4 (at the bottom) all the way to +4 (at the top). There is a vertical line in the middle. The test-takers are on the left side of that line and are arranged according to their ability. Test-takers at the bottom, at -4 logits on the map (identified by the #) were technically the ones with the lowest ability. And at the top, at +4 logits there are 6 test-takers (remember, each # means 2 test-takers, and at the very top were 3 #'s), these were the students with the highest ability. On the right side of the vertical line are items, also arranged according to their difficulty. Item R04 at the bottom, at -3 logits was the easiest, whereas item R17 was the highest item on the map , and hence most difficult.
This distance (in logits) between a test-taker and an item tells us the probability of that test-taker answering that item correctly. So, if the item (in terms of difficulty) is at the same logit of test-taker ability, then there is a 50/50 (50%) probability of answering correctly. If the test-taker measures one logit below the item, so for example if the item is at 0 logits and the test-taker is at -1 logits then, mathematically, the probability of answering that item correctly is .27, or 27%.
How this number is calculated is covered at minute 27:00 in the video.
Clear as mud? :)
@@dr.salehameer thank you soooo much for this detailed explanation. Perfectly clear~:)
Excellent!!!
Thanks a lot :)
Really appreciate it
clear!
Dr Saleh, this is very helpful, would like to ask a question regarding R setting for the maximum score, in the video is stated 9, video 2:39:41 it is assuming that all the score for the questions is the same maximum score? Can it be done for a different range of marks for different question? Say question 1 is score 1-4, question 2, 1 to 6 do i put R as 6?
Hi Ignatius, awfully sorry for the late reply. I would suggest you do a partial credit model where each item has its own rating scale defined. So your specification file would look like this:
Models=
?,?,#,R6
We specified R6 because that was the highest possible score in your example.
I discuss partial credit models in the last part of the video at 03:00:48
Let me know if this was useful.
Nice view
Thanks for this. Very long, but very useful. I have a q: how is a skill like writing unidimensional? Surely it is so complex that it is multidimensional!
You're welcome. Good Q. We need to distinguish between psychological unidimensionality/multidimensionality and psychometric (or measurement) unidimensionality. You are right about writing (or any language ability for that matter) being multidimensional in terms of psychology (e.g., many psychological processes are involved when we write), but in pure measurement terms if these multiple psychological processes are used in unison, the assumption of unidimensionality will hold. See McNamara (1996) and Barkaoui (2014) (both in the reference list on the video) for more.They both discuss this specific issue in greater detail.
We want more videos please about testing or whatever you like?
I plan on that soon. Thanks :)
Dr. Saleh Ameer so welcome 🙏
when's your birthday ?Dr I don’t why I thought it was last January first?
December :)
Why?
Dr. Saleh Ameer lol 😆
It Was close enough am 28 December Capricorn btw