CAUTION! PLEASE READ BEFORE PROCEEDING One more suggestion from my side, please use 0 to 4 scores for the Likert scale when you enter data in the Excel sheet. This will give you more accurate values compared to the scores entered with values 1 to 5. For example, by using 1 to 5 scores, if you have upper group mean as 5 and lower group mean as 1, this is 100% discrimination, so your item discrimination value should be 1. But with this formula, it will be 5 minus 1 and the result divided by 5, that is 4 divided by 5 which equals .80 which is not an accurate value. This problem can be solved if you use a 0 to 4 rating for 5 point Likert scale. In this case, with reference to the earlier example, your upper group mean will be 4 instead of 5 and your lower group mean will be 0 instead of 1. Which will give you 4 minus 0, which is 4, and 4 will be divided by the maximum score possible on one item which is 4. Therefore, 4 divided by 4 will equal 1.0, that is 100% item discrimination value! Hope this helps you. All the best 👍
@@Mar-lo4vm Hi, reference for that formula is Metsämuuronen J (2018). Generalized discrimination index and latent item difficulty. Hope this help. Thanks for watching
Visit the channel for more on psychology, psychometrics, psychological statistics, and research methodology. Subscribe to provide 'positive reinforcement' and press the 'bell' icon to express your 'unconditional' support!
Item discrimination index of a likert scale jo ap bta rahe hai woh kis formula se calculate kar rhe hai? Aur yeh formula kisne diya hai? Research mein hume justification dene hota hai k kisne formula diya hai.
Thanks for watching. Pl subscribe the channel if you find content helpful. We cannot compute item difficulty for Likert scale. Reason is simple, item difficulty is calculated for knowledge test and we don't use Likert scale for knowledge test. And thus, item difficulty is only conceptually and methodologically correct for dichotomous items measuring knowledge including MCQs. I hope this will help. If you have any questions, please do not hesitate to mail to me at shivanandthorat.yt@gmail.com
Hello sir my corrected item-Total correlation for 7 items is. 572,.811,.838,.702,.805,.722,.622 and cronbach alpha for these 7 items is.907 what to do in this matter will it created problem
For discrimination index of a test or likert scale as calculated in this video, You may use the reference table given by Ebel and Frisbie, 1986 for selecting and rejecting an item. For manual calculation of item analysis using the basic formula incorporating RU, RL, N for both item discrimination and item difficulty. You may use the criteria for accepting or rejecting an item given by Noll (1965). Discrimination index using t- Test you may simply use t-table for selecting and rejecting the item.
Sir my scale is 3 points ordinal ( Always=3, Sometimes =2, Never=1) . As ordinal cannot use parametric statistics. which statistics can I use to do item analysis?
Hi, You can use the same procedure that I have used. 3 point Likert scale as you have mentioned can be considered as continuous. Hope this helps. -Shivanand
thank you so much, sir. I am using 4 point scale (SA,A,D,SD) for the SCALE. is there any problem with 4 point LIKERT scale, after item analysis(ITEM DISCRIMINATION METHOD) can I do EFA?
Hi Gyanendra, you can use 4 point likert scale. No problem with that. Just make sure you have scores as 0,1,2,3. If they are 1,2,3,4 then while dividing, devide by maximum possible score on one item minus 1, that is 4 - 1, that is divide by 3
And for descriptive questions, this method is not applicable. You can try other qualitative methods. Content validity ratio (CVR) can be calculated for descriptive questions validation
Hi Sowmya Shree, If responses are in words, you can use find and replace function in excel by selecting each item scores and replacing words by numbers. Follow an appropriate scoring procedure mentioned in the manual of the test you are using. Negative items are generally coded in reverse order. Once all the responses are converted into numbers, use the method mentioned in the video to carry out item analysis. Hope this helps. Best, Shivanand
Hello Thorat, I want to compute item analysis of 5 point likert scale through SPSS. My sample size is 25. Can same process will be used for such small sample size? To standardize a likert scale, should i have to do reliability test after item analysis? There is a item analysis process, where after selecting 27% of upper group and 27% of lower group respondents then further process continues.. I do not remember the actual name of that process.. Do you have any idea about that particular process?
Hi Rumati Das, you can use same procedure. Make sure to read video description before computing. To standardize this test, reliability, validity, and norms are required. Item analysis procedure you are talking about is called as 'item discrimination method'
Item discrimination procedure is already demonstrated in this video. Please watch. Hope this helps. All the best Feel free to write to me in case of further queries.
Hello sir, Sir, I have used three point likert scale for my tool. Is it possible to apply the same corrected item correlation method for the item analysis? Also, my supervisor has suggested to read a book, 'Techniques of Attitude Scale Construction' by Allen L. Edwards. Have you also referred that book ever?
Item difficulty is taken into consideration when you deal with truly dichotomous items (having either right or wrong answer). That also is one of the method used for item analysis. Usually recommended to take items having difficulty value close to 0.5. But again it depends on the objectives of your test. Method for that if you ask, it is the same that I have explained in the very first analysis (item discrimination method for dichotomous data), only cutoff points for retention of the items will vary depending upon the purpose of the test
The discrimination index for Likert scale and dichotomous scale is interpreted in the same way as follow: 0.40 & Above - Very discriminating/Very good item 0.30 to 0.39 - Discriminating/Good item 0.20 to 0.29 - Moderately discriminating item 0.10 to 0.19 - Not discriminating/Marginal item (needs revision) Below 0.10 - Poor/Questionable item Interpretation for correlation coefficients in case of the corrected item-total correlation method (point-biserial) is different. Values of corrected item-total correlation should range from 0.3 to 0.7 for straight coded items and 0.2 to 0.7 for negatively/reverse coded items.
The method that is used in the item discrimination method has no specific person/statistician devising it. But it is simply the procedure followed for item discrimination method of item analysis using Kelley's suggestion (1939) of top and bottom 27% of cases. Anastasi (1988) argue that both practices (methods based on internal consistency like "corrected item-total correlation" and methods like item discrimination [item difficulty in case of dichotomous data]) are important. Whereas, Allen and Yell (1979) argue that both the practices cannot be used simultaneously, that a test constructor must choose one or the other.
Thanks for the video! What if certain items are negative-statement and the other items are positive-statement, what shall I do to decide elimination of item ?
Hi Wingyung Lau, thanks for watching. It doesn't matter whether the item is positive or reverse scored (negative). All you need to be cautious about it is scoring the items carefully. Be it from 1 to 5 (positive item on 5 point Likert scale) or 5 to 1 (negative item on 5 point Likert scale). Criteria to retain the item remains the same. That is, discrimination should be greater than 0.2. If it is less than 0.2 for a particular item, that item should be dropped if it is theoretically not very important. If it is important, then you need to modify it with help of experts and respondents. Good structural changes in the item can improve its discrimination capacity. Hope this helps. Also, there are more videos on channel related to psychometrics and research. Consider subscribing to the channel if you find content helpful.
@@ShivanandThorat Thanks sir! May I know if item-total correlation matrix assumes unidimensonality (i.e. all items are suppose correlate in single direction) and assuming one factor behind all items? Thanks !!
@@ShivanandThorat Also about scoring the items, conceptually my questionnaire items are about "Stage of change" , which all items are conceptually divided into 5 subscales and specifically stage 1 items are negative statement. So do you suggest scoring stage 1 items from 5 (disagree) to 1 (agree); whereas other positive items rate from 1(disagree) to 5 (agree)? im not sure if i got your meaining right. Thanks again :)
Hi Wingyung Lau, you need to use factor analysis to check unidimensionality. It cannot be determined just by looking at all items correlating in one direction
Hello Sir, I want to find discrimination index for 5 point likert scale items and want to apply your calculation. But, confusion is that is there any formula for the same calculation suggested by any eminent researcher? Please response soon as well...
Hello Dhaval sir, I have shown how to compute item discrimination index for the 5 point Likert scale in this video from 10:37 timestamp. Item discrimination has been repeatedly described in the psychometric literature (eg. Ebel, 1965; Johnson, 1951; Mosier, 1940; Anastasi, 1976). Now the problem is, usually it is described in terms of dichotomous data only, they call it as passing percentage. But in Likert scale, we generally don't have any passing or failing criteria which makes it difficult to compute passing percentage as a part of the formula mentioned for computing the item discrimination index. This problem can be solved by using simple mathematical conversions which I have shown in this video. Upper group mean minus lower group mean and then the result is divided by the maximum possible score on one item. Now this function exactly similar to the discrimination formula given in various psychometric books. It's just a conversion used in the Likert type data as you don't have passing percentage in the Likert data. So, for this formula, you can give references of Anastasi and others as mentioned above as this is not a new formula but just a conversion of passing percentage used just like you convert a ratio into percentage.
One more suggestion from my side, please use 0 to 4 scores for the Likert scale when you enter data in the Excel sheet. This will give you more accurate values compared to the scores entered with values 1 to 5. For example, by using 1 to 5 scores, if you have upper group mean as 5 and lower group mean as 1, this is 100% discrimination, so your item discrimination value should be 1. But with this formula, it will be 5 minus 1 and the result divided by 5, that is 4 divided by 5 which equals .80 which is not accurate value. This problem can be solved if you use 0 to 4 rating for 5 point Likert scale. In this case, with reference to the earlier example, your upper group mean will be 4 instead of 5 and your lower group mean will be 0 instead of 1. Which will give you 4 minus 0, which is 4 and 4 will be divided by the maximum score possible on one item which is 4. Therefore, 4 divided by 4 will equal 1.0, that is 100% item discrimination value! Hope this helps you. All the best 👍
@@ShivanandThorat Thanks a lot for responding soon and well. Can I use 0 to 4 likert scale for item discrimination and use 1 to 5 likert scale of the same items for data collection?
CAUTION! PLEASE READ BEFORE PROCEEDING
One more suggestion from my side, please use 0 to 4 scores for the Likert scale when you enter data in the Excel sheet. This will give you more accurate values compared to the scores entered with values 1 to 5.
For example, by using 1 to 5 scores, if you have upper group mean as 5 and lower group mean as 1, this is 100% discrimination, so your item discrimination value should be 1. But with this formula, it will be 5 minus 1 and the result divided by 5, that is 4 divided by 5 which equals .80 which is not an accurate value.
This problem can be solved if you use a 0 to 4 rating for 5 point Likert scale. In this case, with reference to the earlier example, your upper group mean will be 4 instead of 5 and your lower group mean will be 0 instead of 1. Which will give you 4 minus 0, which is 4, and 4 will be divided by the maximum score possible on one item which is 4. Therefore, 4 divided by 4 will equal 1.0, that is 100% item discrimination value!
Hope this helps you. All the best 👍
Thank you for the video. I have one question, what papers is the calculation based on for the dichtomous data?
@@Mar-lo4vm Hi, reference for that formula is Metsämuuronen J (2018). Generalized discrimination index and latent item difficulty.
Hope this help.
Thanks for watching
@@ShivanandThorat Thank you!
@@ShivanandThorat have u got any videos on regression analysis in spss
@@MajidAli-yw3uw Sadly no. I will try and upload video on regression analysis
Very well explained 🎉
Thanks
Thank you very much sir for your deepest guidance in the item analysis.
Visit the channel for more on psychology, psychometrics, psychological statistics, and research methodology.
Subscribe to provide 'positive reinforcement' and press the 'bell' icon to express your 'unconditional' support!
Very well explained 👍
Really informative and very helpful
Nicely presented and explained in simple way
Thanks alot sir
Thank you, Asim
Very clear and helpful video! Thank you so much for uploading :)
Thank you for watching. Subscribe the channel if you find content helpful.
Keep it up , sir👍
Item discrimination index of a likert scale jo ap bta rahe hai woh kis formula se calculate kar rhe hai? Aur yeh formula kisne diya hai?
Research mein hume justification dene hota hai k kisne formula diya hai.
your video is simple to understand thank you , you have not included item difficulty for Likert scale. I am looking for that.
Thanks for watching. Pl subscribe the channel if you find content helpful.
We cannot compute item difficulty for Likert scale. Reason is simple, item difficulty is calculated for knowledge test and we don't use Likert scale for knowledge test. And thus, item difficulty is only conceptually and methodologically correct for dichotomous items measuring knowledge including MCQs.
I hope this will help. If you have any questions, please do not hesitate to mail to me at shivanandthorat.yt@gmail.com
Well explained.
sir plz take a class about what is a dichotomous data, when to use biserial corelation and point biserial corelation
Sure. I will try to cover that in my upcoming videos.
Consider subscribing the channel for more.
Thanks for watching
Useful. Thank you.
very useful video
Pl read pinned comment before undertaking these computations (especially if your data is Likert). Thanks for watching.
Thanks a lot sir
Nice explanation
Thank you! ☺
fruitful
Also I wanted to know do you give classes any other mode rather than RUclips I wanted to join
No ma'am, I don't. But you can have my email id where you can mail your queries shivanandthorat.yt@gmail.com
Very nice exhibition. May you share something on Validity check also please?
Thank you for watching. Will upload more videos soon
Hello sir my corrected item-Total correlation for 7 items is. 572,.811,.838,.702,.805,.722,.622 and cronbach alpha for these 7 items is.907 what to do in this matter will it created problem
Go ahead. No need to delete items as values are just slightly above 0.8
How to select the criteria to discriminate the items ..to select or reject any items
For discrimination index of a test or likert scale as calculated in this video, You may use the reference table given by Ebel and Frisbie, 1986 for selecting and rejecting an item.
For manual calculation of item analysis using the basic formula incorporating RU, RL, N for both item discrimination and item difficulty. You may use the criteria for accepting or rejecting an item given by Noll (1965).
Discrimination index using t- Test you may simply use t-table for selecting and rejecting the item.
Sir plz tell us how discriminating power will be calculated for likert scale items
Dear madam, it is already mentioned in the video. It ranges from 0 to 1. The higher the discrimination index, the higher the power of discrimination.
Sir, please tell how to calculate t test for every item in 5 pt. Likert scale
Reverted to your email. Please check
Sir my scale is 3 points ordinal ( Always=3, Sometimes =2, Never=1) . As ordinal cannot use parametric statistics. which statistics can I use to do item analysis?
Hi,
You can use the same procedure that I have used. 3 point Likert scale as you have mentioned can be considered as continuous.
Hope this helps.
-Shivanand
SIR what is the reference of value should be .3 to .7 pls reply.
According to Cristobal et al. (2007), the items with corrected item-total correlation lower than 0.30 are not acceptable
Streiner, Norman, and Cairney (2015, p. 84) recommend that corrected item-total correlations should lie between .20 and .70.
So combining both, 0.3 to 0.7 is acceptable range. However, for reverse coded items, you can have range from 0.2 to 0.7
Hope this helps. All the best.
Subscribe to the channel if you find content helpful. Feel free to share with your contacts
Good video, my assignment due today. Help. Amazing wow such. Invest dogecoin
Can we have blank or 0 in a likert scale?
Yes, you can have a Likert scale having zero as a score in it. For example, a five-point Likert scale can have scores like 0, 1, 2, 3, and 4.
thank you so much, sir. I am using 4 point scale (SA,A,D,SD) for the SCALE. is there any problem with 4 point LIKERT scale,
after item analysis(ITEM DISCRIMINATION METHOD) can I do EFA?
Hi Gyanendra, you can use 4 point likert scale. No problem with that. Just make sure you have scores as 0,1,2,3. If they are 1,2,3,4 then while dividing, devide by maximum possible score on one item minus 1, that is 4 - 1, that is divide by 3
And you can use EFA or CFA after item analysis
please I have 4 points Likert scale I divided the ID by 3?
Yes
Thanku sir for such a nice explanation. Please suggest the SPSS version which can be downloaded free.
Thank you for watching, madam. Free trial version is available on the IBM website.
@@ShivanandThorat thanku sir.
How to use this for descriptive questions or numerical problem teat
For numerical problem, you can use item difficulty index
And for descriptive questions, this method is not applicable. You can try other qualitative methods. Content validity ratio (CVR) can be calculated for descriptive questions validation
Sir how to find t value for each for item analysis
t value? I didn't get your question
Sir I got response as sA to SD but how to calculate both positive items and negatives in excel sheet
Hi Sowmya Shree,
If responses are in words, you can use find and replace function in excel by selecting each item scores and replacing words by numbers.
Follow an appropriate scoring procedure mentioned in the manual of the test you are using. Negative items are generally coded in reverse order.
Once all the responses are converted into numbers, use the method mentioned in the video to carry out item analysis.
Hope this helps.
Best,
Shivanand
Thanks sir
Hello Thorat, I want to compute item analysis of 5 point likert scale through SPSS.
My sample size is 25. Can same process will be used for such small sample size?
To standardize a likert scale, should i have to do reliability test after item analysis?
There is a item analysis process, where after selecting 27% of upper group and 27% of lower group respondents then further process continues.. I do not remember the actual name of that process.. Do you have any idea about that particular process?
Hi Rumati Das, you can use same procedure. Make sure to read video description before computing.
To standardize this test, reliability, validity, and norms are required.
Item analysis procedure you are talking about is called as 'item discrimination method'
Item discrimination procedure is already demonstrated in this video. Please watch.
Hope this helps. All the best
Feel free to write to me in case of further queries.
Hello sir,
Sir, I have used three point likert scale for my tool. Is it possible to apply the same corrected item correlation method for the item analysis?
Also, my supervisor has suggested to read a book, 'Techniques of Attitude Scale Construction' by Allen L. Edwards. Have you also referred that book ever?
Yes. Corrected item total correlation can be applied. No idea about the book. I haven't referred it so far.
Thanks for watching
@@ShivanandThorat thank you for the prompt response sir. Can I mail you if I face any further problem regarding the same?
@@vandanadaksh6619 sure. shivanandthorat.yt@gmail.com
Sir, can we select items only based on discrimination index? What about item difficulty?
Item difficulty is taken into consideration when you deal with truly dichotomous items (having either right or wrong answer). That also is one of the method used for item analysis. Usually recommended to take items having difficulty value close to 0.5. But again it depends on the objectives of your test. Method for that if you ask, it is the same that I have explained in the very first analysis (item discrimination method for dichotomous data), only cutoff points for retention of the items will vary depending upon the purpose of the test
I hope you found this helpful
If the sample size is 125 what will be the number of upper and lower respondent?
27% of 125 is 33.75.
You can take 34 respondents in upper and 34 respondents in lower group.
Thank you so much🙏💐
For diactomus data 0-1 can use well for discrimination index. As you shown. But I maintained that i will use likert scale but I'm using now diactomus.
how can we interpret DI for Likert scale?
The discrimination index for Likert scale and dichotomous scale is interpreted in the same way as follow:
0.40 & Above - Very discriminating/Very good item
0.30 to 0.39 - Discriminating/Good item
0.20 to 0.29 - Moderately discriminating item
0.10 to 0.19 - Not discriminating/Marginal item (needs revision)
Below 0.10 - Poor/Questionable item
Interpretation for correlation coefficients in case of the corrected item-total correlation method (point-biserial) is different. Values of corrected item-total correlation should range from 0.3 to 0.7 for straight coded items and 0.2 to 0.7 for negatively/reverse coded items.
Sir, can you plz tell the name of the method which is used for likert type? Is it Anastasia's method?
The method that is used in the item discrimination method has no specific person/statistician devising it. But it is simply the procedure followed for item discrimination method of item analysis using Kelley's suggestion (1939) of top and bottom 27% of cases.
Anastasi (1988) argue that both practices (methods based on internal consistency like "corrected item-total correlation" and methods like item discrimination [item difficulty in case of dichotomous data]) are important.
Whereas, Allen and Yell (1979) argue that both the practices cannot be used simultaneously, that a test constructor must choose one or the other.
And its name is simply "item discrimination method"
@@ShivanandThorat Thank you sir..
Sir pls share pdf for likert scale
Which pdf you want?
Thanks for the video! What if certain items are negative-statement and the other items are positive-statement, what shall I do to decide elimination of item ?
Hi Wingyung Lau, thanks for watching.
It doesn't matter whether the item is positive or reverse scored (negative). All you need to be cautious about it is scoring the items carefully. Be it from 1 to 5 (positive item on 5 point Likert scale) or 5 to 1 (negative item on 5 point Likert scale). Criteria to retain the item remains the same. That is, discrimination should be greater than 0.2. If it is less than 0.2 for a particular item, that item should be dropped if it is theoretically not very important. If it is important, then you need to modify it with help of experts and respondents. Good structural changes in the item can improve its discrimination capacity. Hope this helps.
Also, there are more videos on channel related to psychometrics and research. Consider subscribing to the channel if you find content helpful.
Also, read video description carefully before carrying out item analysis. There is slight correction in the method.
@@ShivanandThorat Thanks sir! May I know if item-total correlation matrix assumes unidimensonality (i.e. all items are suppose correlate in single direction) and assuming one factor behind all items? Thanks !!
@@ShivanandThorat Also about scoring the items, conceptually my questionnaire items are about "Stage of change" , which all items are conceptually divided into 5 subscales and specifically stage 1 items are negative statement. So do you suggest scoring stage 1 items from 5 (disagree) to 1 (agree); whereas other positive items rate from 1(disagree) to 5 (agree)? im not sure if i got your meaining right. Thanks again :)
Hi Wingyung Lau, you need to use factor analysis to check unidimensionality. It cannot be determined just by looking at all items correlating in one direction
Hello Sir, I want to find discrimination index for 5 point likert scale items and want to apply your calculation. But, confusion is that is there any formula for the same calculation suggested by any eminent researcher? Please response soon as well...
Hello Dhaval sir,
I have shown how to compute item discrimination index for the 5 point Likert scale in this video from 10:37 timestamp.
Item discrimination has been repeatedly described in the psychometric literature (eg. Ebel, 1965; Johnson, 1951; Mosier, 1940; Anastasi, 1976). Now the problem is, usually it is described in terms of dichotomous data only, they call it as passing percentage. But in Likert scale, we generally don't have any passing or failing criteria which makes it difficult to compute passing percentage as a part of the formula mentioned for computing the item discrimination index. This problem can be solved by using simple mathematical conversions which I have shown in this video. Upper group mean minus lower group mean and then the result is divided by the maximum possible score on one item. Now this function exactly similar to the discrimination formula given in various psychometric books. It's just a conversion used in the Likert type data as you don't have passing percentage in the Likert data. So, for this formula, you can give references of Anastasi and others as mentioned above as this is not a new formula but just a conversion of passing percentage used just like you convert a ratio into percentage.
One more suggestion from my side, please use 0 to 4 scores for the Likert scale when you enter data in the Excel sheet. This will give you more accurate values compared to the scores entered with values 1 to 5.
For example, by using 1 to 5 scores, if you have upper group mean as 5 and lower group mean as 1, this is 100% discrimination, so your item discrimination value should be 1. But with this formula, it will be 5 minus 1 and the result divided by 5, that is 4 divided by 5 which equals .80 which is not accurate value.
This problem can be solved if you use 0 to 4 rating for 5 point Likert scale. In this case, with reference to the earlier example, your upper group mean will be 4 instead of 5 and your lower group mean will be 0 instead of 1. Which will give you 4 minus 0, which is 4 and 4 will be divided by the maximum score possible on one item which is 4. Therefore, 4 divided by 4 will equal 1.0, that is 100% item discrimination value!
Hope this helps you. All the best 👍
@@ShivanandThorat Thanks a lot for responding soon and well. Can I use 0 to 4 likert scale for item discrimination and use 1 to 5 likert scale of the same items for data collection?
@@dr.dhavalvyas1785 Yes yes. You can do that 👍🏻
@@ShivanandThorat Hi, I used a 1-5 Likert scale but getting DI of even above 1 e.g. 1.2, 1.8... is the obvious or where am I going wrong
Sir can you pls share in any pdf of this.pls.if yousay yes then i can share my email id