Thank you so much for these videos, i wish there was more. I used winsteps for rasch analysis in my thesis and had no idea what says these outputs. I understood the parts you explained well but there are still lots of things i need to know. I will research more on google now. Thank you again.
Very helpful video! It would be great if you could do one that shows how to do a Primary Component Analysis of Residuals after running a MFRM. Keep up the good work!
With MFRM, it is pretty tricky to check for unidimensionality, because there is often too much missing data, which makes principal component analysis overwhelmingly impractical.
@@VahidAryadoust right, because a full matrix is required for a PCA. Thinking about it now it seems that the papers I was reading that did use PCA did have the full matrix, but this must be unusual for many studies. Thanks for the reply!
Am using Bond&Fox and its seems the instructions and output are quiet different from winsteps. it has make me little confuse but your video is helpful. can u help
Just an update: Our comprehensive review of Rasch measurement in language assessment has been published: journals.sagepub.com/eprint/YXCQB3NYCICYVFQ4TSCG/full
Hope your time allows for answering my question, I have implemented Rasch analysis and calculated ROC, Sensitivity, Specificity, PPV & NPV, to evaluate ASD screener validity. The data analysis results of Sensitivity, Specificity support the validity of the screener, while the Rasch analysis results don’t. Could you please give me your interpretation of this result, how to combine these results? Sensitivity (correctly identify autistic child) 0.85 high. Specificity (false positives) 0.99 very high. NPV 0.99 very high. PPV 0.02 very low.
Thanks for your question. I am not quite sure if ROC, Sensitivity, Specificity, PPV & NPV are used in the context of the Rasch model. These are used in classification and data-mining methodologies such as CART and neural networks etc. You may find this paper useful: www.researchgate.net/publication/265249052_CaMLA_Working_Papers_Predicting_Listening_Item_Difficulty_with_Language_Complexity_Measures_A_Comparative_Data_Mining_Study_Predicting_Listening_Item_Difficulty_with_Language_Complexity_Measures_A_Com
From the menu bar on top of the main interface, choose Output files --> ITEM File IFILE --> Excel --> Ok. You can transfer the excel table into your Word file.
Hi Vahid. Thank you for putting up part 2. It heIped clarify some issues with Winsteps unidimensionality analysis I ran in an effort to investigate the extent of local item dependence in a test with a problematic clustering format. Could I ask for some advice?
Thanks, Vahid. I am conducting a unidimensionality analysis in order to uncover the extent of Local Item Dependence (LID) in a test with a problematic 6x3 cluster format (3 items x 6 options). I suspect LID to be present and to manifest itself as another dimension. I am aware of the following areas in the Winsteps unidimensionality output tables that need to be checked: Table 23.0 Raw var explained by measures: obs vs expected [big difference suggests non-unidimensionality] Unexplained variances in 1st to 5th contrasts: >2 Table 23.99 largest standardized residual correlations: >0.5 or 0.6 …. >0.7 highly locally dependent However, apart from these, I am not sure how to read the other graphs and tables in the 50-page printout. Are there other places in this analysis that can provide more evidence of LID between items? Many thanks. Nigel ================ Here is a sample of the printout of my data so far. I am only at the beginning of data collection and I will need to get a few 100 more samples, so I am aware that the analysis is already skewed. Table 23.0 TABLE 23.0 volev 3 levels ONLY data.67ss.2018.01.x ZOU020WS.TXT Mar 20 2018 8:52 INPUT: 67 PERSON 90 ITEM REPORTED: 67 PERSON 90 ITEM 2 CATS WINSTEPS 4.0.1 -------------------------------------------------------------------------------- Table of STANDARDIZED RESIDUAL variance in Eigenvalue units = ITEM information units Eigenvalue Observed Expected Total raw variance in observations = 140.0860 100.0% 100.0% Raw variance explained by measures = 56.0860 40.0% 39.7% Raw variance explained by persons = 23.0105 16.4% 16.3% Raw Variance explained by items = 33.0755 23.6% 23.4% Raw unexplained variance (total) = 84.0000 60.0% 100.0% 60.3% Unexplned variance in 1st contrast = 5.7106 4.1% 6.8% Unexplned variance in 2nd contrast = 5.1860 3.7% 6.2% Unexplned variance in 3rd contrast = 4.2206 3.0% 5.0% Unexplned variance in 4th contrast = 3.6635 2.6% 4.4% Unexplned variance in 5th contrast = 3.4410 2.5% 4.1% ….. TABLE 23.99 vlt 3 levels ONLY data.67ss.2018.01. ZOU020WS.TXT Mar 20 2018 8:52 INPUT: 67 PERSON 90 ITEM REPORTED: 67 PERSON 90 ITEM 2 CATS WINSTEPS 4.0.1 -------------------------------------------------------------------------------- LARGEST STANDARDIZED RESIDUAL CORRELATIONS USED TO IDENTIFY DEPENDENT ITEM ------------------------------- |CORREL-| ENTRY | ENTRY | | ATION|NUMBER IT |NUMBER IT | |-------+----------+----------| | .75 | 23 23 | 52 52 | | .74 | 7 7 | 9 9 | | .72 | 23 23 | 24 24 | | .69 | 7 7 | 37 37 | | .65 | 1 1 | 38 38 | | .61 | 9 9 | 30 30 | | .61 | 2 2 | 52 52 | | .57 | 3 3 | 4 4 | | .57 | 22 22 | 86 86 | | .56 | 10 10 | 11 11 | | .54 | 10 10 | 16 16 | | .52 | 25 25 | 47 47 | | .51 | 24 24 | 52 52 | | .50 | 4 4 | 56 56 | | .48 | 40 40 | 79 79 | |-------+----------+----------| | -.66 | 7 7 | 68 68 | | -.60 | 7 7 | 60 60 | | -.50 | 9 9 | 68 68 | | -.49 | 32 32 | 77 77 | | -.48 | 30 30 | 81 81 | -------------------------------
Nigel, from what I see, Table 23.0 indicates there is some degree of multidimensionality, as indicated by Unexplned variance in 1st contrast = 5.7106 (there are around 5.7 items (rounded up to 6) which cluster together. Next go to TABLE 23.2, and find out what these 6 items are. Also, maybe find out what these items have in common and decide whether you have a "good" secondary dimension (still relevant to your construct) or a "bad" one (irrelevant to the construct being measured. I would say you won't need to examine all tables. Table TABLE 23.99 presents "Q3" statistics. These are correlations between residuals. This shows if items might be locally dependent. Q3 is commonly used in IRT (both unidimensional and multidimensional models such as DINO, DINA, etc.). I can see several items are locally dependent such as 23 and 52; 7 and 9. etc. Do you need to keep all these items in your instrument? What do they measure? For further information regarding Q3, i recommend the following papers: Chen, W. H., & Thissen, D. (1997). Local dependence indexes for item pairs using item response theory. Journal of Educational and Behavioral Statistics, 22, 265-289. Yen, W. M. (1984). Effects of local item dependence on the fit and equating performance of the three-parameter logistic model. Applied Psychological Measurement, 8, 125-145. My paper on measuring listening will also be published soon. If you are interested, you can leave you email address. I will send you a copy. hope it helps.
Hi Vahid, Many thanks for the reply. It was very helpful. I will take a look at the papers you suggested, and I would also like to read yours, too. My email is ndaly@hotmail.com. But I still have a question or two about the data set, if you don’t mind. Table 23.2 shows that items A (7), B (9), C (37), D (14), E (80) and F (4) "cluster together" (see Table below). I have two questions: 1. In Table 23.99, the only items in common with Table 23.2 are 7 and 9. The Q3 statistic indicates they are highly correlated (.74). In know this is very technical, but unless I am mistaken, both Tables indicate items that may have dimensionality issues, so what do the different statistical results mean? 2. Relatedly, can items A-F possible indicate local item dependence? Unlike the correlations in Table 23.99, Table 23.2 indicates only (problematic) individual items. So, does this mean I only need to qualitatively examine these items and cross-check them with each other if I am looking for items that are locally dependent? Or with other items? I am expecting to find several item clusters (every 3-item cluster share the same 6-answer option set), but only 7 and 9 are in the same 3-item cluster. TABLE 23.2 vle 3 levels ONLY data.67ss.2018.01.x ZOU020WS.TXT Mar 20 2018 8:52 INPUT: 67 PERSON 90 ITEM REPORTED: 67 PERSON 90 ITEM 2 CATS WINSTEPS 4.0.1 -------------------------------------------------------------------------------- CONTRAST 1 FROM PRINCIPAL COMPONENT ANALYSIS STANDARDIZED RESIDUAL LOADINGS FOR ITEM (SORTED BY LOADING) ---------------------------------------------- |CON- | | INFIT OUTFIT| ENTRY | | TRAST|LOADING|MEASURE MNSQ MNSQ |NUMBER IT | |------+-------+-------------------+----------| | 1 1 | .76 | -3.20 .97 .29 |A 7 7 | | 1 1 | .70 | -2.47 .89 .25 |B 9 9 | | 1 1 | .59 | -2.47 .93 .30 |C 37 37 | | 1 1 | .51 | -1.71 .91 .41 |D 14 14 | | 1 1 | .48 | -.71 .71 .40 |E 80 80 | | 1 1 | .44 | -1.45 .86 .86 |F 4 4 | | 1 1 | .33 | -1.04 .97 2.03 |G 1 1 | | 1 1 | .32 | -.71 .90 1.37 |H 38 38 | | 1 1 | .32 | -.44 .96 .94 |I 56 56 | | 1 2 | .27 | -.20 .80 .57 |J 88 88 | Thanks again and I hope you can make more videos like the tutorials you posted. There are surprising few Rasch tutorials on UA-cam, especially given that it has become such a widely used analytic tool. All the best. Nigel
Nigel, regarding your first question, TABLE 23.2 and TABLE 23.99 present different things. As you know the unidimensionality analysis extracts the dimensions in Rasch residuals and in your analysis, Table 23.0 shows that you have 5 potential dimensions in the residuals.. TABLE 23.2 above shows the items that load on the first dimension (or contrast). For example, item A (7) is loading on dimension 1 (also known as CONTRAST 1 FROM PRINCIPAL COMPONENT ANALYSIS) by .76; thus this item is likely to be testing something else in addition what you intend it to be testing (maybe a construct-irrelevant concept or something else.). You can find factor loading of items on the second dimension in TABLE 23.12 etc. You can interpret them in the same way. On the other hand, Table 23.99 indicates potential local dependence (not unidimensionality). For example, items 23 and 52 in your analysis are highly dependent (their loading is 0.75), so you might want to investigate if one of these items might have to be deleted.. maybe look at their content and see what makes them look similar or whether answering one of them depends on answering the other one--in which case you will need to eliminate this dependence. Hope it helps.
@@learn5081 it is called "interaction" analysis or "bias" analysis. You can find the codes and relevant information from here: www.winsteps.com/a/ftutorial3.pdf
First of all, thanks for the video. I have a question though: I ran my analysis on Winstep and checked table 23. However I can't find the table in the end with Item residual correlations to assess local item dependence. I wonder if it's because new versions of Winstep put this table somewhere else, but I can't seem to find it. Do you know where I can find it? Do you know if there's any other way to get that information with R or MPlus? Thanks :)
how can i calculate the score of a rasch-based scale (of which validity was verified via rasch analysis) via rasch analysis? (what I mean by the "score" is like the "mean" which can be used for various analyzes like t tests or ANOVA in the SPSS). And how can i perform analysis with that rasch-based score?
If I understand your question well, you can transfer the measures from Winsteps to SPSS and run follow-up analysis there. Go to the 'output files' menu in Winsteps --> 'PERSON file PFILE=' --> Choose excel or SPSS --> OK. You will get a number of different variables in the SPSS or excel file generated, including person measures.
@@VahidAryadoust Thank you so much for answering me. Are those person measures "raw scores" or "logits"? I am confused with these words and the word "anchoring". I guess I need to make some kind of transformation on those person measures so I can run follow-up analysis and correlate those measures with other scaled scores
Thank you for the detailed interpretation, please continue with
Fit Statistics, and Targeting.
Thank you so much for these videos, i wish there was more. I used winsteps for rasch analysis in my thesis and had no idea what says these outputs. I understood the parts you explained well but there are still lots of things i need to know. I will research more on google now. Thank you again.
You are welcome. I will make more videos soon as I find some free time. All the best
Very helpful video! It would be great if you could do one that shows how to do a Primary Component Analysis of Residuals after running a MFRM. Keep up the good work!
With MFRM, it is pretty tricky to check for unidimensionality, because there is often too much missing data, which makes principal component analysis overwhelmingly impractical.
@@VahidAryadoust right, because a full matrix is required for a PCA. Thinking about it now it seems that the papers I was reading that did use PCA did have the full matrix, but this must be unusual for many studies. Thanks for the reply!
Am using Bond&Fox and its seems the instructions and output are quiet different from winsteps. it has make me little confuse but your video is helpful. can u help
I am happy to help, if you have any question.
thanks a lot. it's really helpful.
Just an update: Our comprehensive review of Rasch measurement in language assessment has been published:
journals.sagepub.com/eprint/YXCQB3NYCICYVFQ4TSCG/full
GOOD FORMATIVE PRESENTATION, IF YOU JUST GO SLOWLY :)
Hope your time allows for answering my question, I have implemented Rasch analysis and calculated ROC, Sensitivity, Specificity, PPV & NPV, to evaluate ASD screener validity. The data analysis results of Sensitivity, Specificity support the validity of the screener, while the Rasch analysis results don’t.
Could you please give me your interpretation of this result, how to combine these results?
Sensitivity (correctly identify autistic child) 0.85 high.
Specificity (false positives) 0.99 very high.
NPV 0.99 very high.
PPV 0.02 very low.
Thanks for your question. I am not quite sure if ROC, Sensitivity, Specificity, PPV & NPV are used in the context of the Rasch model. These are used in classification and data-mining methodologies such as CART and neural networks etc.
You may find this paper useful:
www.researchgate.net/publication/265249052_CaMLA_Working_Papers_Predicting_Listening_Item_Difficulty_with_Language_Complexity_Measures_A_Comparative_Data_Mining_Study_Predicting_Listening_Item_Difficulty_with_Language_Complexity_Measures_A_Com
Thanks DR. please how to export Winsteps file specially tables to word document
From the menu bar on top of the main interface, choose Output files --> ITEM File IFILE --> Excel --> Ok.
You can transfer the excel table into your Word file.
@@VahidAryadoust Thank you so much Dear DR.
Hi Vahid. Thank you for putting up part 2. It heIped clarify some issues with Winsteps unidimensionality analysis I ran in an effort to investigate the extent of local item dependence in a test with a problematic clustering format.
Could I ask for some advice?
Sure, Nigel; i am happy to help, if I can.
Thanks, Vahid. I am conducting a unidimensionality analysis in order to uncover the extent of Local Item Dependence (LID) in a test with a problematic 6x3 cluster format (3 items x 6 options). I suspect LID to be present and to manifest itself as another dimension.
I am aware of the following areas in the Winsteps unidimensionality output tables that need to be checked:
Table 23.0
Raw var explained by measures: obs vs expected [big difference suggests non-unidimensionality]
Unexplained variances in 1st to 5th contrasts: >2
Table 23.99 largest standardized residual correlations:
>0.5 or 0.6 …. >0.7 highly locally dependent
However, apart from these, I am not sure how to read the other graphs and tables in the 50-page printout. Are there other places in this analysis that can provide more evidence of LID between items?
Many thanks.
Nigel
================
Here is a sample of the printout of my data so far. I am only at the beginning of data collection and I will need to get a few 100 more samples, so I am aware that the analysis is already skewed.
Table 23.0
TABLE 23.0 volev 3 levels ONLY data.67ss.2018.01.x ZOU020WS.TXT Mar 20 2018 8:52
INPUT: 67 PERSON 90 ITEM REPORTED: 67 PERSON 90 ITEM 2 CATS WINSTEPS 4.0.1
--------------------------------------------------------------------------------
Table of STANDARDIZED RESIDUAL variance in Eigenvalue units = ITEM information units
Eigenvalue Observed Expected
Total raw variance in observations = 140.0860 100.0% 100.0%
Raw variance explained by measures = 56.0860 40.0% 39.7%
Raw variance explained by persons = 23.0105 16.4% 16.3%
Raw Variance explained by items = 33.0755 23.6% 23.4%
Raw unexplained variance (total) = 84.0000 60.0% 100.0% 60.3%
Unexplned variance in 1st contrast = 5.7106 4.1% 6.8%
Unexplned variance in 2nd contrast = 5.1860 3.7% 6.2%
Unexplned variance in 3rd contrast = 4.2206 3.0% 5.0%
Unexplned variance in 4th contrast = 3.6635 2.6% 4.4%
Unexplned variance in 5th contrast = 3.4410 2.5% 4.1%
…..
TABLE 23.99 vlt 3 levels ONLY data.67ss.2018.01. ZOU020WS.TXT Mar 20 2018 8:52
INPUT: 67 PERSON 90 ITEM REPORTED: 67 PERSON 90 ITEM 2 CATS WINSTEPS 4.0.1
--------------------------------------------------------------------------------
LARGEST STANDARDIZED RESIDUAL CORRELATIONS
USED TO IDENTIFY DEPENDENT ITEM
-------------------------------
|CORREL-| ENTRY | ENTRY |
| ATION|NUMBER IT |NUMBER IT |
|-------+----------+----------|
| .75 | 23 23 | 52 52 |
| .74 | 7 7 | 9 9 |
| .72 | 23 23 | 24 24 |
| .69 | 7 7 | 37 37 |
| .65 | 1 1 | 38 38 |
| .61 | 9 9 | 30 30 |
| .61 | 2 2 | 52 52 |
| .57 | 3 3 | 4 4 |
| .57 | 22 22 | 86 86 |
| .56 | 10 10 | 11 11 |
| .54 | 10 10 | 16 16 |
| .52 | 25 25 | 47 47 |
| .51 | 24 24 | 52 52 |
| .50 | 4 4 | 56 56 |
| .48 | 40 40 | 79 79 |
|-------+----------+----------|
| -.66 | 7 7 | 68 68 |
| -.60 | 7 7 | 60 60 |
| -.50 | 9 9 | 68 68 |
| -.49 | 32 32 | 77 77 |
| -.48 | 30 30 | 81 81 |
-------------------------------
Nigel, from what I see, Table 23.0 indicates there is some degree of multidimensionality, as indicated by Unexplned variance in 1st contrast = 5.7106 (there are around 5.7 items (rounded up to 6) which cluster together. Next go to TABLE 23.2, and find out what these 6 items are. Also, maybe find out what these items have in common and decide whether you have a "good" secondary dimension (still relevant to your construct) or a "bad" one (irrelevant to the construct being measured. I would say you won't need to examine all tables.
Table TABLE 23.99 presents "Q3" statistics. These are correlations between residuals. This shows if items might be locally dependent. Q3 is commonly used in IRT (both unidimensional and multidimensional models such as DINO, DINA, etc.). I can see several items are locally dependent such as 23 and 52; 7 and 9. etc. Do you need to keep all these items in your instrument? What do they measure?
For further information regarding Q3, i recommend the following papers:
Chen, W. H., & Thissen, D. (1997). Local dependence indexes for item pairs using item response theory. Journal of Educational and Behavioral Statistics, 22, 265-289.
Yen, W. M. (1984). Effects of local item dependence on the fit and equating performance of the three-parameter logistic model. Applied Psychological Measurement, 8, 125-145.
My paper on measuring listening will also be published soon. If you are interested, you can leave you email address. I will send you a copy.
hope it helps.
Hi Vahid,
Many thanks for the reply. It was very helpful. I will take a look at the papers you suggested, and I would also like to read yours, too. My email is ndaly@hotmail.com.
But I still have a question or two about the data set, if you don’t mind.
Table 23.2 shows that items A (7), B (9), C (37), D (14), E (80) and F (4) "cluster together" (see Table below). I have two questions:
1. In Table 23.99, the only items in common with Table 23.2 are 7 and 9. The Q3 statistic indicates they are highly correlated (.74). In know this is very technical, but unless I am mistaken, both Tables indicate items that may have dimensionality issues, so what do the different statistical results mean?
2. Relatedly, can items A-F possible indicate local item dependence? Unlike the correlations in Table 23.99, Table 23.2 indicates only (problematic) individual items. So, does this mean I only need to qualitatively examine these items and cross-check them with each other if I am looking for items that are locally dependent? Or with other items? I am expecting to find several item clusters (every 3-item cluster share the same 6-answer option set), but only 7 and 9 are in the same 3-item cluster.
TABLE 23.2 vle 3 levels ONLY data.67ss.2018.01.x ZOU020WS.TXT Mar 20 2018 8:52
INPUT: 67 PERSON 90 ITEM REPORTED: 67 PERSON 90 ITEM 2 CATS WINSTEPS 4.0.1
--------------------------------------------------------------------------------
CONTRAST 1 FROM PRINCIPAL COMPONENT ANALYSIS
STANDARDIZED RESIDUAL LOADINGS FOR ITEM (SORTED BY LOADING)
----------------------------------------------
|CON- | | INFIT OUTFIT| ENTRY |
| TRAST|LOADING|MEASURE MNSQ MNSQ |NUMBER IT |
|------+-------+-------------------+----------|
| 1 1 | .76 | -3.20 .97 .29 |A 7 7 |
| 1 1 | .70 | -2.47 .89 .25 |B 9 9 |
| 1 1 | .59 | -2.47 .93 .30 |C 37 37 |
| 1 1 | .51 | -1.71 .91 .41 |D 14 14 |
| 1 1 | .48 | -.71 .71 .40 |E 80 80 |
| 1 1 | .44 | -1.45 .86 .86 |F 4 4 |
| 1 1 | .33 | -1.04 .97 2.03 |G 1 1 |
| 1 1 | .32 | -.71 .90 1.37 |H 38 38 |
| 1 1 | .32 | -.44 .96 .94 |I 56 56 |
| 1 2 | .27 | -.20 .80 .57 |J 88 88 |
Thanks again and I hope you can make more videos like the tutorials you posted. There are surprising few Rasch tutorials on UA-cam, especially given that it has become such a widely used analytic tool.
All the best.
Nigel
Nigel, regarding your first question, TABLE 23.2 and TABLE 23.99 present different things. As you know the unidimensionality analysis extracts the dimensions in Rasch residuals and in your analysis, Table 23.0 shows that you have 5 potential dimensions in the residuals.. TABLE 23.2 above shows the items that load on the first dimension (or contrast). For example, item A (7) is loading on dimension 1 (also known as CONTRAST 1 FROM PRINCIPAL COMPONENT ANALYSIS) by .76; thus this item is likely to be testing something else in addition what you intend it to be testing (maybe a construct-irrelevant concept or something else.). You can find factor loading of items on the second dimension in TABLE 23.12 etc. You can interpret them in the same way.
On the other hand, Table 23.99 indicates potential local dependence (not unidimensionality). For example, items 23 and 52 in your analysis are highly dependent (their loading is 0.75), so you might want to investigate if one of these items might have to be deleted.. maybe look at their content and see what makes them look similar or whether answering one of them depends on answering the other one--in which case you will need to eliminate this dependence. Hope it helps.
Can we perform the analysis in FACETS?
C HS you can perform a similar analysis in facets, but the output you will get is limited, compared with what Winsteps provides.
@@VahidAryadoust thx for the reply. May I ask the code for performing the similar analysis in FACETS?
@@learn5081 it is called "interaction" analysis or "bias" analysis. You can find the codes and relevant information from here: www.winsteps.com/a/ftutorial3.pdf
Really have not see where the last table came from
What table are you referring to?
First of all, thanks for the video. I have a question though: I ran my analysis on Winstep and checked table 23. However I can't find the table in the end with Item residual correlations to assess local item dependence. I wonder if it's because new versions of Winstep put this table somewhere else, but I can't seem to find it. Do you know where I can find it? Do you know if there's any other way to get that information with R or MPlus? Thanks :)
This is probably due to the very low correlation between residuals.
You can use the ltm or MIRT R package to run similar analysis.
@@VahidAryadoust Thanks, I'll try
Can you share rasch calculator?
Do you mean the software?
You can download it from: www.winsteps.com/ministep.htm
how can i calculate the score of a rasch-based scale (of which validity was verified via rasch analysis) via rasch analysis? (what I mean by the "score" is like the "mean" which can be used for various analyzes like t tests or ANOVA in the SPSS). And how can i perform analysis with that rasch-based score?
If I understand your question well, you can transfer the measures from Winsteps to SPSS and run follow-up analysis there. Go to the 'output files' menu in Winsteps --> 'PERSON file PFILE=' --> Choose excel or SPSS --> OK. You will get a number of different variables in the SPSS or excel file generated, including person measures.
@@VahidAryadoust Thank you so much for answering me. Are those person measures "raw scores" or "logits"? I am confused with these words and the word "anchoring". I guess I need to make some kind of transformation on those person measures so I can run follow-up analysis and correlate those measures with other scaled scores
@@Gazi_UNLU You will get to see both measures and raw scores. The measures are in 'logits". You do not have to make any transformation on these.
@@VahidAryadoust Thank you very much, i got it.
@@Gazi_UNLU welcome and best wishes.