tag:blogger.com,1999:blog-3721871707048712457.post2868874220260253184..comments2024-08-13T00:28:36.247-04:00Comments on Breaking BI: Conducting a 2-sample Z Test in TableauBreaking BIhttp://www.blogger.com/profile/02551920506874509998noreply@blogger.comBlogger8125tag:blogger.com,1999:blog-3721871707048712457.post-32837275823536943602014-06-13T09:45:27.786-04:002014-06-13T09:45:27.786-04:00That's an interesting question. Unfortunately...That's an interesting question. Unfortunately, I don't have access to that email address at the moment. Changing the color shouldn't be such a difficult task if you understand Compute Using. Perhaps this post will help<br /><br />http://breaking-bi.blogspot.com/2013/07/introduction-to-table-calculations.html<br /><br />Thanks,<br /><br />Brad LlewellynBreaking BIhttps://www.blogger.com/profile/02551920506874509998noreply@blogger.comtag:blogger.com,1999:blog-3721871707048712457.post-16094559463294350342014-06-13T02:44:10.549-04:002014-06-13T02:44:10.549-04:00Brad,
Thanks a ton for the response! And it worke...Brad,<br /><br />Thanks a ton for the response! And it worked perfectly. I need help on just one more doubt which arose on solving this problem.<br /><br />Now, as we are working in the client environment.and the client doesn't have Tableau 8.1 neither does he has R. We took help from the blog below to perform t-test<br />http://community.tableausoftware.com/docs/DOC-1428<br /><br />As mentioned earlier, the initial problem was to provide t-test result (p-value) even when a user is making changes in filter selections. We have solved this issue now. <br />The difficulty which we are facing (and I think we might face even after integrating Tableau with R) is using p-value to colour the bar chart. To understand the issue better please have a look at the tableau file which I have shared with you on your official email.<br /><br />The workbook attached has two sheets <br />o T-test table<br />o Desired chart<br />- Using the aforementioned blog we are able to perform t-test (in t-test table). The p-value is getting changed if we change any/all of these filters (problem 1 solved)<br />- But the client requirement is to colour the chart (aggregated at retailer name level) in ‘desired chart’ using p-value. Eg. - Significant (p<0.05) as green and non-significant as red<br />- As p-value is a table calculation, we are struggling to change colour of bar graph. <br />- Any guidance on this would be extremely helpful<br /><br />Thanks,<br />Surya<br />Suryahttps://www.blogger.com/profile/16432223034965376304noreply@blogger.comtag:blogger.com,1999:blog-3721871707048712457.post-92035082742143630792014-06-12T14:39:13.489-04:002014-06-12T14:39:13.489-04:00Surya,
Take note of the beginning of the post whe...Surya,<br /><br />Take note of the beginning of the post where we create a calculation using<br /><br />SUM(<br /> IF YEAR( [Order Date] ) = 2012 THEN [Profit] END<br />)<br /><br />You could replace that IF statement with anything you want. You could even create a parameter to allow the user to select the value via a dropdown! On a different note, Tableau 8.1 has an integration with R that will allow you to do this type of thing much easier.<br /><br />Thanks,<br /><br />Brad LlewellynBreaking BIhttps://www.blogger.com/profile/02551920506874509998noreply@blogger.comtag:blogger.com,1999:blog-3721871707048712457.post-4280515452389015462014-06-12T09:59:22.678-04:002014-06-12T09:59:22.678-04:00Brad, Thanks for the awesome blog. I was wondering...Brad, Thanks for the awesome blog. I was wondering if Tableau would allow us to do these tests at different levels with the same data set. Now, the data is at customer level and say, we add one more categorical variable like Location (North, South being two values). <br /><br />So, I need Z statistic for the following pairs:<br />2011 Profit of North Vs 2012 Profit of North<br />2011 Profit of South Vs 2012 Profit of South<br />2011 Profit Vs 2012 Profit<br /><br />*I'll have a drop down to select North and South.<br /><br />Appreciate your help! Expecting a reply soon!<br /><br />Thanks,<br />SuryaSuryahttps://www.blogger.com/profile/16432223034965376304noreply@blogger.comtag:blogger.com,1999:blog-3721871707048712457.post-77332211479387431532013-06-21T08:13:12.964-04:002013-06-21T08:13:12.964-04:00It's simply a method of "proving" a ...It's simply a method of "proving" a concept. Anyone can look at a graph and see something. The underlying question is always, "Is that really true?" Statistics gives you a quantifiable measure to determine whether or not you are seeing what you think you are seeing. Most business users won't have any use for statistics. You would have to find an audience that is ready for the next level of analytics, and these audiences are not very common.Breaking BIhttps://www.blogger.com/profile/02551920506874509998noreply@blogger.comtag:blogger.com,1999:blog-3721871707048712457.post-35027673436755144342013-06-21T02:25:14.380-04:002013-06-21T02:25:14.380-04:00Yes. Bit clear. Imagine i new to world of statisti...Yes. Bit clear. Imagine i new to world of statistics and histograms.<br />How well it brings decision making capabilities or should i use these kind of graphs for decision making.<br /><br />Why not comparative analysis or my basic bar and line or 23 types of tableau charts. I want to understand its core and basic use. Can you help ? <br /><br />I studied a lot of statistics, p value, normal distribution etc. where does it really used in real life and how it will be benefitted to clients. from your exp, i wish to know.Anonymoushttps://www.blogger.com/profile/18056058130432023678noreply@blogger.comtag:blogger.com,1999:blog-3721871707048712457.post-21134823871493228822013-06-20T09:32:12.912-04:002013-06-20T09:32:12.912-04:00Harihara,
That description is as simple as I feel...Harihara,<br /><br />That description is as simple as I feel comfortable saying. Some people might say<br /><br />"Since the P-value is extremely high, the means are the same."<br /><br />These people would be wrong. Alas, I digress.<br /><br />The P-value is the first thing taught in an introductory statistics course, which is now mandatory in many universities. If your business users don't understand them, then it's probably best not to use the procedure at all. If they really want to know whether the distributions look the same, but don't understand P-values, just show them a histogram and let them decide that way.<br /><br />I apologize if this sounds "high and mighty", but I don't feel comfortable spreading misinformation.<br /><br />Thanks,<br /><br />Brad Llewellyn<br />Associate Consultant<br />Mariner, LLC<br />brad.llewellyn@mariner-usa.com<br />http://www.linkedin.com/in/bradllewellyn<br />http://breaking-bi.blogspot.comBreaking BIhttps://www.blogger.com/profile/02551920506874509998noreply@blogger.comtag:blogger.com,1999:blog-3721871707048712457.post-83764559412807586702013-06-20T08:39:14.624-04:002013-06-20T08:39:14.624-04:00Brad,
I wish to know how to explain use of these ...Brad,<br /><br />I wish to know how to explain use of these statistical test to laymen ? Can you pls let me know bit more of below line ?<br /><br />As we suspected, the P-value is extremely high, meaning that we don't have enough evidence that say that these two samples come from populations with different means.<br />Anonymoushttps://www.blogger.com/profile/18056058130432023678noreply@blogger.com