i have developed a scale, i want to know how to score responses of each item. what is the procedure to decide whether we have to start from 0 or 1 for each item.
Before you develop the scale, you need a range of numerical values which could range from 1-5. This range gives the weight of the responses. For example, if the total respondents is 30 and the scale range is 1= strongly disagree, 2= disagree, 3= some how agree, 4= agree and 5= strongly agree and from you data 3 persons strongly disagree, you will then have 3x1=3, 8persons disagree(8x2=16), 5 persons some how agree(5x3=15), 9 persons agree(9x4=36) and 5persons strongly agree(5x5=25). total score = 3+16+15+36+25=95. Points=95/30=3.167. You can conclude that the respondents some how agree to the item because it falls within the range of some how agree
I have attached a useful paper by two experts in survey question writing.
In terms of how you compute your ultimate scale score, it makes no difference whether your items scores begin with 1 or 0, so pick whichever one will make most sense for the specific kinds of questions that you are writing.
The scoring to start from zero or one is not a problem. the important issue is the scale validity and reliabiltiy. the items should measure what it supposed to measure in the content domain and according to the construct you are interested in and the objective of the scale.
usually Likert scale is summed to get its total score to reflect the level of score achievment.
Since you have developed your own, unique, first-time used index, then scoring it should be relatively easy. Once you have assigned numerical values for the values of each variable, you simply add them up. Until you have amassed a large amount of data, doing anything more 'fancy' statistical wise, would be inappropriate.
Before you develop the scale, you need a range of numerical values which could range from 1-5. This range gives the weight of the responses. For example, if the total respondents is 30 and the scale range is 1= strongly disagree, 2= disagree, 3= some how agree, 4= agree and 5= strongly agree and from you data 3 persons strongly disagree, you will then have 3x1=3, 8persons disagree(8x2=16), 5 persons some how agree(5x3=15), 9 persons agree(9x4=36) and 5persons strongly agree(5x5=25). total score = 3+16+15+36+25=95. Points=95/30=3.167. You can conclude that the respondents some how agree to the item because it falls within the range of some how agree
The level of each item is determined by the following formula: (highest point in Likert scale − lowest point in Likert scale)/the number of the levels used .For instance 7 point likert scale will be (7 − 1)/7 = 0.80, where 1 - 1.80 reflected by “very low”, 1.81 - 2.60 reflected by “low”, 2.61 - 3.40 reflected by “moderate”, 3.41 - 4.20 reflected by “high”, and 4.21 - 5 reflected by “very high”. ( depending on whether you rated 1 as lowest and 7 highest)
Hello dear JUDITH NGULI, can u please give any reference for using this formula? Actually i am also using a Likert scale and want to use this formula but i need an authentic justification to use it in my thesis... thanku :)
Hi....
Scoring start from 0 or 1 is not a big issue. U can follow 5 point Likert scale. To prevent neutral response each statement should have five points ‘strongly agree’ , ‘agree’ , ‘undecided’, ‘disagree’ and ‘strongly disagree’ and may be scored as 5,4,3,2,1.
Judith, kindly review the example. The calculation is for 5 point Likert scale and not for 7.
I am curious, though--is it acceptable practice to assign numeric values (1-5) to a 5-level Likert scale *after* the responses have been collected? In other words, the respondents were simply given a choice of five levels (strongly agree, somewhat agree, neutral, somewhat disagree, strongly disagree), but they did not use a 1-5 score. Can the researcher assign these values later in order to carry out analysis?
I have a question that follows this one. Assuming that it does follow all the rules of validity and reliability. You have made a Likert Scale questionnaire and you use 1-5; there are 6 questions. You add all the 1's. the 4's, the 5's you divide the results by the amount of questions (5 questions and you scored a 20 b/c that is the sum of all of the 5 questions) What does the Mean tell you? (it would be 4 in this case and NOT using SPSS) This is not a high level questionnaire it is more like do you like this product or not.
If confusion arises from this issue of assigning weights to a likert scale (Strongly Agree, Agree, etc.) and using it to interpret result, I think better use mode. Do not use the mean because mode is the most appropriate. The weight assigned is just arbitrary.
Scoring start from 0 or 1 is not a big issue but it can confuse a lot in understanding the results and interpretation.
Let
0= strongly disagree,
1= disagree,
2=some how agree,
3= agree and
4= strongly agree
So you are using 5 point Likert scale and you start coding from 0 but your summary results (e.g. averages) never approach to 5 because you have given the maximum number 4. It will take a while to understand that Five (5) point Likert Scale is gauging upto 4. While a layman or reader is expecting to maximum gauge 5 for"5 point Likert Scale".
ON the other hand if you start coding from 1 then it will be easy to understand because the summary result of 5 point Likert Scale is gauging upto 5.
So the name and number matches. No clarifications. No Confusion.
Cronbach's alpha assumes that all of the items are simply summed (i.e., given equal weights, rather than the item-weighted scores produced by a factor analysis). I suggest you calculate a simple sum and check its correlation with the scale you produced via factor analysis. If the correlation is above .9 or so, I would stick with the simpler version.
It is possible that your factor score is a "standardized variable" with a mean of 0 and a standard deviation of 1. You can check that with descriptive statistics.
Thanks for the comments David. Yes it is actually a standardized variable. The items I had used for the PCA were on different likert scales and it was suggested that I perform a z-score before doing the PCA. Was Wrong with this? I am totally new to all this and would just want to create a scale for one of my independent variables.
Now that I have this scale (-7 to 7) and this is after doing a total sum of the items for my cases. Are these the values to be used in my regression, or I am still few steps away to obtain my final scores . I was trying to measure uncertainty avoidance which has not have a universal agreeably scale in the literature. But it is generally categories into low, middle and high uncertainty avoidance behaviours.
Most factor analysis procedures use correlation matrices, so standardizing the variables makes no difference. And the whole point of computing a score is so that you can use that scale in subsequent analyses, such as a regression analysis. If you think it would
When I tried to find out about standardization, I saw mixed conclusions. Hence, ended up deciding to standardized to be safer, especially that I have different scales for the items. I think I meant to ask, what kind of likert scale suits my scale range (-7 to 7). Or can I go with the low, medium and high measure of uncertainty avoidance.
For my study I used a 5-Likert scale, 1= strongly agree , 2= agree, 3=neutral, 4 disagree and 5= strongly disagree, I have computed the sum of the items related to each variable. When analysing I got this table, how can I interpret the results.
Statistics
Facilitating conditions
N Valid 65
Missing 0
Mean 23.1692
Median 24.0000
Mode 28.00
Kurtosis -.353
Std. Error of Kurtosis .586
Sum 1506.00
you need to develop a scale base on your likert Scale range, I recommend you read Ngodigha S.A article on likert scale it will surely guide you.
I have used three point likert scale from 1 to 3 , 3 being good , 2 for fair and 1 for poor. How to assign individual weight so that i can calculated index value to rank?
Your scale will be 3-1/3 = 0.67
0.67 - 1.67 = Poor
1.68 - 2.35 = Fair
2.34 - 3.00 Good
Your scale will be 3-1/3 = 0.67
0.67 - 1.67 = Poor
1.68 - 2.35 = Fair
2.34 - 3.00 Good
Dear All
I used 4-point likert scale in my study. These below are Mean and SD for sub-factors of a scale.
Mean Standard Deviation
3.06 0.441
3.10 0.499
2.35 0.767
2.86 0.594
3.13 0.493
These Above are the mean and standard deviation of sub-factors of a scale. One of my expert commented and asked from me about these mean values which is given below:
(How were these mean scores obtained? By averaging the responses to the constituent items? How did these scores vary? Between 1 and 4?)
Dears
Yes this is true I got these values by averaging the responses to the constituent items. But Dears I could not understand commenter's next sentence. Kindly anyone who tells me about this sentence. How did these scores vary? Between 1 and 4?
Dear All
I used 4-point likert scale in my study. These below are Mean and SD for sub-factors of a scale.
Mean Standard Deviation
3.06 0.441
3.10 0.499
2.35 0.767
2.86 0.594
3.13 0.493
These Above are the mean and standard deviation of sub-factors of a scale. One of my expert commented and asked from me about these mean values which is given below:
(How were these mean scores obtained? By averaging the responses to the constituent items? How did these scores vary? Between 1 and 4?)
Dears
Yes this is true I got these values by averaging the responses to the constituent items. But Dears I could not understand commenter's next sentence. Kindly anyone who tells me about this sentence. How did these scores vary? Between 1 and 4?
Can any body reply this question.
https://www.researchgate.net/post/Should_we_assign_3_or_0_for_No_opinion_in_a_five-point_Likert_scale
Whatever the scale is , the law for evaluating the result is :
Interval width = (No. of responses-1)/No. of results
for example :
Likert Scale : 5 categories (from 1-5)
if you want 3 results : Poor , Moderate , Good
Interval width = (5-1)/ 3 = 4/3 = 1.33
Then :
Poor : 1 - 2.33
Moderate : 2.34 - 3.67
Good > 3.67
Alh Abubakar Sadiq Sani
Ali Al-fahham
Do you have a reference for this?
thanks
@Ammanuel Lemma, thank s for the PDF which brings out clearly how to interpret Likert scale data.
Ok pls note that when you give a value of Likert scale to each of your categories,then they are just labels.Or should I say ordinal labels.So you could give a 1 or 5 to start with,just that you move in order which is 12345 or 54321.Till teh time you remember while analysis that the 1 you named was so and so variabel ,it is fine to label it in any order
Dear Ali Al-Fahham,
Can you please help us a reference for the mentioned formula:
Interval width = (No. of responses-1)/No. of results
That formula is actually pretty meaningless, because all it does is (in the example given) is transform the width from 1.0 to 1.3, but it leaves the width the same across all the categories. As long as you have equal widths, you have the same ordinal scoring as you did when you started.
I would like to thanks all researchers.Your answers and PPT helping me a lot to understand what likert scale is and how to use it.
Hi everyone how to compute descriptives for likert data where the responses are not assigned a numeric value
would you pls support me on how to analyze the5 point likert scale data or send me the reference materials
Dedefo Teshite If you have multiple items that each measure the same basic concept, then you may be able to combine them into a scale that will be interval-level. If you have a single item, then you should use non-parametric statistics.
Ali, that is a very good answer to the question. what i need is a reference to justify it. for example if we
say 3.47 is good, who do we quote as a reference? Thank you
I am also struggling with interpreting Likert scale data of scales 1-5 and 1-10. Someone shared with me these categories ranges 1-2.5 = low 2.6-3.5= neutral and 3.6-5=high for 5-point scale while 1-4= low, 4.1-6=neutral and 6.1-10= high for the 10 point scale. Is this correct and there a reference to it?
If you need to combine different Likert-scorings into the same scale, the most widely accepted approach its to standardize each variable (i.e. mean of 0 and standard deviation of 1), before you add them together.
Even so, I would do an Exploratory Factor Analysis first, to make sure you do not get a "methods factor" that separates the two sets of response categories.
Philomena Ngugi please can you give me the reference for your answer because that is what exactly I want to do for my research but I failed to find the reference for it. I am still searching for the reference.
Hi
For the benefit of future questions
if you have a questionnaire the answer can be obtained via the statistic program, it is very easy.
same Sabina Alatari Ngodigha
Total standard scores= 5 * 30 pers= 150
Total scores for answers= 95 (see; Sabina Alatari Ngodigha )
The ratio of answers to standard scores= 95/150=0.633
Score ranging= 0.633*5= 3.167
but, if you want to know the strength of a group of independent variables such as (Score ranging) on a dependent variable. for this problem, I suggest that the test Factor loadings of exploratory analysis, it will be very good for ur Quest.
Please note that there is not statistical basis for this kind of procedure. In particular, arithmetic operations such as multiplication and division do not apply to ordinal data such as Likert-scored items.
Probabilities are based on the frequencies at which things occur, not on responses to attitude items.
Using the following formula:
Percentile value =[((Sum of actual scoring scale-Sum of Min. of scoring scale))/(Range of Sum of scoring scale)]*100%
How to Assign the weightage of score for each level? I would like to say if very important is assigned as 5 or 1?
Likert-scored variables are ordinal, which means they are simply ordered, with no way to know the "weight" attached to each level. For example, you know that 1 is less than 2, but not how large this difference is. Similarly, you know that 4 is less than 5, but there is no way to know if this gap is the same as the difference between 1 an 2.
Hi, I am very curious as it seems there have been two camps on saying likert can be weighted and another camp say can't
So I am wondering, then how do I measure using all those SPSS software etc
Pardon my ignorance as I am trying to find out whether my respondents agree whether training is crucial in preventing accidents etc
The scale (Likert) is based on the expected degrees of possible or possible occurring results and according to what the statistical experiment has achieved. This came from the essence of defining the statistical experiment, which is any experiment whose expected results can be determined before they happen. For example, when teaching a group of students in a class that includes thirty students and the limitation in the delivery of the subject was to depend on the level of students ’ability to receive the subject and previous information was provided that the percentage of students with low ability is not more than one tenth of the total students and those with high ability are not more than One-tenth of the total number of students, and eight-tenths of students are of medium ability, so the scale will take the value [1, 8, 1]. So we will have the expected values are [3, 24, 3]
The scale (Likert) is based on the expected degrees of possible or possible occurring results and according to what the statistical experiment has achieved. This came from the essence of defining the statistical experiment, which is any experiment whose expected results can be determined before they happen. For example, when teaching a group of students in a class that includes thirty students and the limitation in the delivery of the subject was to depend on the level of students ’ability to receive the subject and previous information was provided that the percentage of students with low ability is not more than one tenth of the total students and those with high ability are not more than One-tenth of the total number of students, and eight-tenths of students are of medium ability, so the scale will take the value [1, 8, 1]. So we will have the expected values are [3, 24, 3]
This scale does not measure variables absolutely. So I think, it would be better to assign 1 for the lowest response.
@Dilli Raj Pant: we know that Likert scale is ordnial. Now assigning a specific numeric to loweat or highest csnt be fixed. Because lowest/ highest are qualitative expression to which a research is assigning a numeric value. In some case lowest number will be hight for an investigator and in some it will be low. Only one rule shall govern i.e. numeric values of all questions shall be represented in same manner.
For the "Yes" or "No" option scale, the scoring can be done as providing weightage as 0 for No and 1 for each correct answer or you can define the weightage as decided by you.
Imagine that you are in a class consisting of 30 students, and from a prior information that the number of students with high thinking ability is around 1/8, and those with low ability at the same previous level and the rest are of moderate ability ... What is the appropriate scale? Inevitably the scale will be based on the expected results and in light of the previous information it will be 1, 8, 1 for high, moderate, and low, and accordingly the expected values will be 3, 24, 3. From the previous example, we conclude that the score scales are completely dependent on the expected results, and this inevitably requires in-depth knowledge of the nature of the phenomenon.
That starting from the number one or with the number zero, this depends on the fact that the minimum response represents the two types of zeros, namely the sampling zero and the structural zero respectively. In the case where the zero is relatively like the expression never, then the number one will be the limit that represents the response while when it is absolutely zero like The expression never too, so the number is zero that represents the mentioned response.
A “Likert scale” is the sum of responses to several Likert items. These items are usually displayed with a visual aid, such as a series of radio buttons or a horizontal bar representing a simple scale.
A “Likert item” is a statement that the respondent is asked to evaluate in a survey. In the example below, the information, “The checkout process was easy,” is a Likert item. The table as a whole is the Likert scale.
A Likert scale-
1-3 or 1-5 or 1-7 or 0-100. It depends on the respondents and the sampling technique.
I have one article related to a Likert scale.
https://www.tandfonline.com/doi/full/10.1080/21580103.2020.1862712?src=
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design