Hyphothesis Testing for Hold Vs AHT
Six Sigma – iSixSigma › Forums › Old Forums › Software/IT › Hyphothesis Testing for Hold Vs AHT
 This topic has 13 replies, 8 voices, and was last updated 12 years, 3 months ago by Srinivasan.

AuthorPosts

May 23, 2006 at 6:13 am #26473
Hi, Need some help in figuring out which hypothesis test to be used if i have to find out if Hold Time has an impact on Average Handle Time in a call center environment. Both the data sets are continous.
Arun0May 25, 2006 at 4:50 am #64328Arun,
I am a call center SS Black Belt as well. I do what you are speaking of quite often, but am not really using hypothesis test in the strict sense.
Once you have a good ided of your popoulation distributions, (normality assumption) you should also pop the data into a control chart and look for special cause points that our out of spec with the rest of the group.
If your data is normal (or assumed normal) and you have a process that is in control statistically from the control charts. (I usually set up a random sample too for the individual call measurments rather then actual averages) you can use a regression analysis (fitted line in mintab) to take a look at the effects each other have. The idea is for a regression, if you have an Rsq value of around (70% for me rule of thumb) you can state that the ammount of variation in A can be explained for 70% by the variation in B (AHT vs Hold Time)
Once you have established that there is a correlation, you need to dig down and take a look at the residuals:
Residuals vs. Fitted, Distribution of Residuals, look at the best fit for normality and make sure you see a good fit for the residuals. In minitab there is a button for 4 in 1 graphs that will contain all of this when you run the regression.
But now let us look at it from a overview standpoint. Hold time is a component of AHT, Right. The basic formula for AHT takes into account Hold Time. Thus you would have some automatically built in correlation for a certain ammount.
I usually break down the AHT in a Excel Bar Graph that shows all of the different proportions in one bar graph. AHT = ACD time +outbound call time + After Call Work(ACW + etc) it is different based on where you work.
So looking at AHT as the denomintaor, everything else in the numerator, what proportion of AHT is taken by each call, this will give you a stronder understanding of the system I think then a regression at this point. If you do choose to run a regression, grab a stats book to make sure you assume all the proper requirments to determine if the regression is indeed significant
hope this helps.0May 25, 2006 at 6:29 am #64329Thanks a lot Joe for the detailed explanation. I always prefer using regression analysis for comparing the metrics. However, I was trying to get more used to with Hypothesis tests. I was referring to a few tables from various sources and they weren’t clear in terms of which test to be used between the two.
As you rightly mentioned, AHT is a superset of Hold and has got a high probability of correlation. I would like to know if I have to use a H Test which one is appropriate?0May 25, 2006 at 2:07 pm #64330A hypothesis test will only tell you how your populations look compared to each other. Equal, less, greater.
You can use them to determine what kind of population fit you have, normal, weibull, etc..
I am not really sure what good comparing the means and variance of (F and T test) of AHT and Hold.
You can really only say, AHT is (,=) Hold Times using hypothesis tests. This will not give you a good explanation of how one effects the other, just a method of looking at their populations and describing them compared to each other.
Hope this helps,
Joe0May 26, 2006 at 4:47 am #64332Thanks a lot for the excellent piece of input Joe. I was always under the impression that hypothesis testing is used to determine the impact of X over Y.
If you could pls. give me some more examples in which situations does hypothesis tests help. When should one use Hypothesis Tests in a project what is the use of those tests?0May 27, 2006 at 4:25 am #64334
joe musallamParticipant@joemusallam Include @joemusallam in your post and this person will
be notified via email.I recently did a study looking at resason for decreases in incoming call volume, as part of that study I was able to do a ton of testing of hypothesis conclusions and regression tests.
give me your email and I will send you a copy of the report (about 30 pages) and you can use it for an example reference.
Hope I can help,.
Joe0May 28, 2006 at 4:35 am #64335Hey Joe, Thanks a lot for this. You have been of great help!. You can email me the documents to [email protected]
Thanks,0May 31, 2006 at 6:37 am #64337hi joe,
i found your explanation really helpful..im a green belt learner…could be please send me a copy of the same..so i can have a overview
thanx
deep0June 7, 2006 at 8:32 am #64343Hi, I would also Like to look into the hypthesis examples, if you could please also mail me at [email protected], since i am also working on hypthesis implementation in software – Code review
0June 7, 2006 at 12:23 pm #64344Hi Joe,
Can you also mail me your report on the Hypothesis tests. It will be very much useful.
Thank you for the same in advance.
Sridhar0June 7, 2006 at 12:28 pm #64345Hi Joe Musallam ,
Can you also mail me your report on the Hypothesis tests. It will be very much useful.
Sorry….you can send me the file to [email protected]
Thank you for the same in advance.
Sridhar0June 8, 2006 at 12:55 am #64346I am removing any copyright and proprietary info, I will post soon
Joe0June 23, 2007 at 1:40 am #64741What is the young guava leave hyphothesis ?
the young guava leaves has a pectin that can heal a diarhea
0June 12, 2008 at 8:50 am #65106Hi Joe,
I am doing a similar study on AHT .could you please forward it to [email protected]
Thanks
Uma0 
AuthorPosts
The forum ‘Software/IT’ is closed to new topics and replies.