Split-Half Reliability and the Spearman-Brown Coefficient using SPSS

hello this is dr. grande welcome to my video on split half reliability and the Spearman brown coefficient and SPSS in counseling assessment when we're evaluating an instrument we oftentimes look at the reliability of the instrument and reliability is a measure of internal consistency one of the popular reliability statistics is chromebooks alpha however in this video i'll be demonstrating another relatively popular method which is the split half method and specifically the Spearman Brown coefficient so taking a look at these fictitious data I have loaded the data editor you can see I have scores from six items so these would be scores from six items on the same instrument and let's assume that these items are measuring some type of career aptitude so the items would be related to a skill or maybe a few skills related to a specific career so a higher score on one of the items would represent a stronger aptitude to the specific career and a lower score would represent a weaker aptitude and you can see that the minimum value is one and the maximum is ten so we want to assess the reliability for this instrument based on the scores that we have available so we have 20 records and we have the scores for each of the six items so I'm going to go to analyze and then scale and then reliability analysis this is what the dialogue looks like by default you can see the default model is chromebooks alpha which I mentioned earlier I'm going to switch this to split half so the model will be the split half model and for the first example I'm going to load all the items in as they appear here and the order they appear here so hold down control and select items one through six and move them over together so the top item is item one and the bottom is item six under statistics I'm really just interested in the descriptives for scale so that's all I'm going to check off here I'll click continue then okay and you can see I have 20 valid cases no cases were excluded there are no missing values here and the next table has the reliability statistics listed so you can see we have a part one and a part 2 and the chromebooks alpha is returned for each of these parts and you can see here for number of items of course there's three because we had six items total and this is split half so it's going to be three compared to three and notice the superscript here is a and next to the part two number of items value the superscript is B you can see the items are one two and three and four five and six that's because I load them from one to six so it took the top three that I loaded in is comparing them to the bottom three so you can see the chromebooks alpha for part one is 0.9 0 6 and for part two point six nine two and then it returns a correlation between forms and that value is 0.7 zero one now this statistic the correlation between forms can be obtained by taking the average or the sum of the first three items scores and then the second set of three items scores and computing the correlation so if I move here to Excel you can see I have the same fictitious data loaded in this excel worksheet items one through six all the scores so first I'm going to take the average of the first three items so the equal sign average and then I'm going to select item one two and three and then hit enter and then for part two I'm going to take the average of the second set of three scores that'll be four five and six so if I autofill these all the way down for all 20 records and then I move here and take the correlation that bc o r re L in Excel and I put the averages for part one in the first argument and the average is for part two and the second you can see I get 0.75 I were to take the sum instead move over here sum and the first three items and then for part two the sum second three items the second set of three items four five and six and again autofill that all the way down then calculate the correlation here between Part one and Part two I get the same result so moving back to SPSS we can see that the point seven zero one may not be ideal because it's used half the items and compared it to the other half and the Spearman brown coefficient compensates for this so with the Spearman brown coefficient we would be getting a reliability value that theoretically would represent the reliability for the full test and not for half the test but you can see there are two rows here equal length and unequal length in this case the values are identical 0.8 to 4 but they're not always identical we select equal length will interpret this Spearman brown coefficient when the number of items is even and unequal length when the number of items is odd in this case we have six items which is an even number so we interpret equal length 0.82 for this table also provides the Guttman split half coefficient and you can compare that value here 0.8 one for fairly close to the Spearman brown coefficient in this case so what I've done here is not an unusual way to conduct the split half to take the first half of the items in compared to the second half but the most popular way is to compare the odd and even items so I move back to analyze then go back to scale reliability analysis I'm just going to move the items around so the items one three and five are the first three items and two four and six are the second three so this is odd compared to even so the order in which we load the items has an effect on the value that you'll see for the correlation between forms and the Spearman Brown coefficient under statistics of course scales still checked off and that's fine click continue and then click OK and you can see here we have a different result the chromebox alpha for part 1 0.7 4/5 for part two point seven nine one and the correlation between forms 0.8 to nine so again if I go back to excel I have already calculated the part 1 and part 2 sums in this case so if I select this one for example and move back you can see that's one three and five and if we take a look at the part two again just move to the left you can see that's item two item four and item six and I've calculated the sum of the odd and the even items and then taking the correlation just as I did here for these other examples and again we have 0.8 to 9 which matches the correlation between forms here in the SPSS table the Spearman brown coefficient in this case again we're still interpreting equal length is 0.9 0 7 so a bit better when looking at odd compared to even then when looking at the first spri items compared to the second three items and then the Guttman's split half coefficient again is fairly close Oh point nine zero three so if we were to compare this to the value for chromebox alpha we would see that the chromebooks alpha value is very close if I go back to analyze scale reliability analysis and change this to alpha it does not matter what the order of the items would be for chromebox alpha I could have these in any order and it's gonna be the same result I click OK and you can see it's 0.8 7 8 so fairly close to the point nine zero seven and to the point eight two four that we had for the first example I hope you found this video on split half reliability and the Spearman Brown coefficient to be useful as always if you have any questions or concerns feel free to contact me and I'll be happy to assist you

Channel: Dr. Todd Grande

Views: 53,655

Rating: 4.9188404 out of 5

Keywords: SPSS, Excel, Microsoft Excel, Spearman-Brown Coefficient, Spearman, Brown, Spearman-Brown, Cronbach, alpha, Cronbach’s alpha, Gutman, Gutman Split-Half, correlation between forms, correlation, odd items, even items, inter-item, reliability, validity, construct, consistency, inter-item reliability, inter-item consistency, counseling, Grande

Id: GK82PJCncNk

Length: 11min 13sec (673 seconds)

Published: Fri Jan 01 2016