Welcome

What is Real Statistics Using Excel?

Real Statistics Using Excel is a practical guide for how to do statistical analysis in Excel plus free statistics software. This software package extends Excel’s built-in statistical capabilities. This will enable you to more easily perform a wide variety of statistical analyses in Excel.

What does Real Statistics Using Excel consist of?

Real Statistics Using Excel is comprised of the following four components:

Real Statistics Resource Pack

An Excel add-in that extends Excel’s standard statistics capabilities. It provides you with advanced worksheet functions and data analysis tools. This will enable you to more easily perform a wide variety of practical statistical analyses. The software supports Excel 2007, 2010, 2013, 2016, 2019, 2021, and 365 for Windows. It also supports Excel 2011, 2016, 2019, 2021, and 365 for the Mac. The current version is Rel 8.9.1 (released 2 October 2023). There is also limited support for Excel 2002 and 2003.

Real Statistics Website (i.e. this site)

  • Lets you download a free copy of the Real Statistics Resource Pack
  • Provides tutorials and descriptions of how to perform a variety of statistical analyses using built-in Excel capabilities. You will also learn how to use the supplemental capabilities provided by the Real Statistics Resource Pack.
  • Presents numerous examples in the form of Excel worksheets which you can download to your computer

For the student and the novice, the Real Statistics website is an excellent tutorial for learning the basic concepts of statistics and how to do statistical analysis. For all users, it provides a step-by-step guide for how to do statistical analysis in the Excel environment and the tools necessary to carry out these analyses.

You can use this website to learn how to perform statistical analyses in Excel even without using the Real Statistics Resource Pack. But we recommend that you download the resource pack so that you can have access to its powerful capabilities.

Real Statistics Examples Workbooks

Consists of fourteen Excel files that contain all the examples shown on the website. These example files can be downloaded for free (click here). Each example focuses on a specific statistical concept and has been designed to demonstrate simple concepts before moving on to more complicated topics. 

Over time, the webpages on the Real Statistics website are being updated so that you will be able to download an Excel file with any of the examples found on that particular webpage.

Real Statistics Community

Each webpage has a Comments section where you can make suggestions, identify errors, ask questions to others in the community, or request their advice.

You can also get updates about new releases and other information via the @Real1Statistics twitter feed.

How do I get started?

Step 1: If you elect to use the Real Statistics Resource Pack, click on the following icon. You will be given the opportunity to download and install for free the Real Statistics Resource Pack. 

free-download-real-statistics

Once you have downloaded and installed the Real Statistics Resource Pack, you will be able to use the supplemental capabilities with your copy of Excel. This is described throughout the rest of the website and is summarized in Using Real Statistics Functions and Real Statistics Data Analysis Tools.

If you choose not to download the resource pack or examples now, you can do so later at any time.

Step 2: Browse through the website to learn how to perform a wide range of statistical analyses in Excel using standard built-in as well as supplemental Real Statistics capabilities. We suggest that you begin by clicking on the Website Introduction (and especially Organization of the Website). These will explain how to navigate the website to get the information you need to run any specific statistical test or learn about any particular topic.

Why do statistical analysis in Excel?

The reasons for choosing Excel are as follows:

  • It is widely available and many people already know how to use it
  • It is not necessary to incur the cost of yet another tool
  • There is no need to learn new methods of manipulating data and drawing graphs
  • It already contains some basic statistics functions and data analysis tools
  • It is much easier to see what is going on since unlike the more popular statistical analysis tools very little is hidden from the user
  • Excel provides the user with a lot of control and flexibility

This makes Excel an ideal tool for learning statistical concepts and performing some basic statistical analyses. Unfortunately, its built-in statistics capabilities are limited. As a result, many prefer to use statistical tools such as R, SPSS, or SAS for carrying out more advanced statistical analyses.

We created the Real Statistics Resource Pack to address these shortcomings. This software package contains various supplemental tools that enable you to carry out a wide range of advanced statistical analyses without leaving the Excel environment. You can download the Real Statistics Resource Pack free of charge from this website.

853 thoughts on “Welcome”

  1. Hi there,

    What format do i work in Excel in order to analyse the data I have? I am new to research, and have no idea where to start or how best to catorgize my data.

    Thanks

    Reply

  3. Dear Charles, I read really a lot of questions and replies of many users, trying to download to run this very useful statistics tool. However, many of them and including me are suffering by the same behavior of their Excel 2007-issue, asking for password, compiler errors etc. although Solver is switched on. Your answer is everytime referring to the instruction =VER(), in my case it gives 5.11 Excel 2007. Also hints like toggling in options by Alt-TI (which simply does not work in my version because it everytime points to inserting a hyperlink) does not help. I use the Dutch version from Excel but don’t think it is the cause. In Office 365 I succeeded to get it running but on some laptops with installed 2007 it was not succesfull. I’m getting desperate and don’t know how to continue.
    I am a 6Sigma blackbelt and certainly like this program but at this moment stuck like many other people. If you have any other out of the box suggestions, please let us know!

    Reply

  4. I downloaded xrealstat add in for excel 365 and add it in to excel and tried to run a Kaplan meier curve and every time a try I get an error message:

    “Compile error in hidden module:survival.”

    I am not sure how to get this add in to work. Please help.

    Reply

  5. Hi Charles:

    I installed and run successfully the pack but when I have used the pack only appear the first number in all the results. and later I in have to make enter in all the functions to make it run or I have to restart excel. I was working in a university class with 30 people and many of us have the same error.

    What can we do?

    Reply

  6. Hi Charles,

    This is very helpful! 🙂

    But I have a question. I tried comparing SPSS and Real Stat ICC estimates and confidence interval but there were discrepancies in their values. What could be the possible reason for this? I have read somewhere that SPSS uses McGraw and Wong ICC formula. I’m going to use the Real Stat in my study.

    Thank you

    Reply

  8. If we have a test of 50 items with mean of 30 and standard deviation of 6, what will be the reliability of the test using kuder Richardson method?
    1 0.70
    2 0.82
    3 0.68
    4 0.78

    Reply

  10. Questions: the package is “free to download”, but what does it cost to use?

    and what’s the upper limit on Nobs for survival analysis? Does it depend on what version of Excel I’m running?

    Thanks,

    Reply

  12. Hi Charles,

    I’ve only found your website some minutes ago and just wanted to say that not only do your posts seem from a glance to be very extensive and well thought-out, but your commitment to replying to comments is astounding. I couldn’t imagine giving thoughtful answers to all of these commenters over the years, even those whose questions are hard to understand, are in different languages, or have missed points you’ve already explained – especially when so many never reply to clarify their questions.

    Perhaps you could set up a donate link on the front page and get something in return for so much statistical help!

    Thank you for this website. I’m not really a math person, but I can’t help but think that in some small way, you’re making the world better.

    Reply

    • Hello Anon and thank you very much for your very gracious remarks. I have tried to share my knowledge and expertise and since my retirement, I have enjoyed creating and expanding the Real Statistics project.
      I really should add the donation link to the front page. For now, you can find it on the second menu tab
      Donation
      Charles

      Reply

  13. Hi Charles,
    Hoping that you are fine in these days of crisis.
    I am doing a research on PUJ obstruction. Herewin we test DRF (Differential renal function). This DRF is a linear variable ranging from 0 to 50.
    There are two more variables:
    First one: T1/2 (this is also a linear variable ranging from 0.5 to 40)
    Second one: NORA ( This is also a linear variable ranging from 0.03 to 8.54)
    I want to test which one of the two variables (i.e T1/2 or NORA) can predict the DRF in more reliable way.
    Thanking you

    Reply

    • I am well in this period of crisis. I hope that you too are well.
      Assuming the relationship between T1/2 and DRF is linear and the relationship between NORA and DRF is linear, you can build two linear regression models: one between T1/2 and DRF and the other between NORA and DRF and see which performs better.
      If you have a lot of data, you can use most of the data to create the regression models and use the rest to see which of the two regression models performs better. This can be measured in many ways, but the usual approach is to see which model has the lower SSE on the test data.
      If the relationships are not linear, you can build non-linear regression models.
      Charles

      Reply

        • Dear Charles,
          Thanks a lot for your suggestions and advice. As suggested by you. I found a linear relationship between T ½ and DRF and also between NORA and DRF. I was able to do linear regression.
          In addition to these three variables (i.e. T1/2, NORA and DRF) there is another variable: “Need of Surgery” ; meaning thereby which patient has undergone surgery and which patient did not need surgery (Same as Titanic survival analysis in your example) This is a “Yes/No” variable.

          After this I have run a Binary logistic regression taking “Need of surgery” as outcome variable and tested T1/2, NORA and DRF. These three variables were tested individually and together. I found that all these variables have some role in predicting the need of surgery.

          Now I want to create cut of points for these variables individually as to beyond this point the variable predicts need of surgery and between these points the variable does not predict need of surgery confidently and beyond this point the variable tells us that there is no need of surgery. How do I do this?

          Thanks

          Dr. Prashant Mulawkar

          Reply

          • Hello Prashant,
            An interesting question. Since there is an interaction between the three independent variables, there may not be a cutoff for each variable that doesn’t depend on the others. You also need to determine the accuracy criteria to use regarding the need for surgery since the model won’t be 100% accurate. All I can suggest is to experiment with values for the three independent variables to see what accuracy values look satisfactory.
            Charles

  14. Hi, Charles
    I hope you and yours are well during these difficult times.

    I have, what I hope, is a very easy suggestion. You do fairly frequent updates to this package, which is great! However, I was recently trying to decide if I needed an update and was trying to determine exactly what version of RealStats I had installed and could not easily figure it out. I had to go to the update page and compare the features in my release to narrow it down.

    To make answering this question a little easier, I suggest you add an “About This Release” menu item in the “Misc” tab OR add a fifth button on the right side of the dialog box that says “About” or something like that. The information provided should be the version installed and any other stuff you might want to include. You could, for example, also include:

    1) a suggested citation for any researchers that use your tool in their work or,
    2) a request for donations or a link to your website or,
    3) a “check for updates” link or button.

    Thanks, again, for this really great resource. Take care.

    Regards,
    Chris V.

    Reply

  15. In the chapter of the analysis of the historical series I do not see the Garch and the various families of garch. this topic has not been developed?
    anyway congratulations for the site

    Reply

  16. Hi Charles,

    For my bachelor thesis, I’m conducting an event study on the response of the Brazilian market on the 2018 election of Bolsonaro. I compiled a sample of 70 Brazilian firms operating in 8 different industries. I used the mean adjusted return model with an estimation window starting at -200 and ending at -30, and an event window starting at -1 and ending at +1. My goal is to get the average cumulative abnormal return of every individual industry. I calculated this by taking the average of all the individual CARs of companies operating in the respective industries. My problem lies in the significance testing of the average cumulative abnormal return of a specific industry. I did this by using the regular t-test, but I want to check the robustness of my results by applying a non-parametric test. Which tests do you recommend? And is there an Excel spreadsheet available on how to apply that test?

    Thank you in advance,

    Reply

    • Hello Yannieck,
      If the regular t test is the 2 independent sample t test, then the usual non-parametric version is the Mann-Whitney test. See the following webpage on how to perform this test in Excel. The Real Statistics software can be used to perform the test in Excel.
      Mann-Whitney Test
      Charles

      Reply

  17. Dear charles,
    I am only a beginner into doing research projects.I am currently assigned a tasks to assess growth variations of a plant species by comparing the growth and development of that particular plant by investigating the used fibre cells with the new fibre cells.What kind of experimental design do I have to use? and what kind of experimental lay out would that be?
    Thanks.

    Reply

  18. Doctor good evening, again I am very grateful for this powerful tool that makes life easier for us. Doctor, the option: Real Statistics Support for KDE, is it not working, will there be an error or is it my error ?.
    Thank you

    Doctor buenas noches, nuvamente estoy muy agradecido, por esta herramienta poderosa que nos facilita la vida. Doctor, la opción: Real Statistics Support for KDE, no esta funcionando, existirá lagú error o es error mio?.
    Muchas gracias

    Reply

  19. Hi Charles, I am happy to see that the website is still here! I also hope that everything is fine at this hard time in Italy. I need once again some help to understand which statistic is better to use when analyzing some data. I am measuring flow (in ml/min) in blood vessels before to induce a lesion that will block the flow. That is my control group (A) for which I record the time it takes to block the flow. Then I test the effect of a drug (B) into the occlusion time. Sometimes the treatment prolongs, sometimes it decreases this time. Most of the time the effect is concentration dependant: at the highest doses, there is never an occlusion. By never I mean within the predefined time of observation which cannot be infinite nor equal to the lifetime of the animal, of course. So most of the time I have a group A control and then several B1,n… where n is all the different doses of my treatment, and my hypothesis is if the mean of any of this Bn are different from the A control. What is the best way to analyze these kinds of data? I think we used in the past a non-parametric approach but I`m not sure if it is the best. What do you think?

    Reply

    • Hello Alessandro,
      This seems like a fit for ANOVA. If you get a significant result you would use Dunnett’s follow-up test. If the assumptions for ANOVA are not met, then you could use a non-parametric test, such as Kruskal-Wallis, or Welch’s ANOVA, but this depends on which assumptions are not met. If you use Kruskal-Wallis, then Steel’s test is probably the appropriate follow-up test. All of these tests can be found at
      One-way ANOVA
      Charles

      Reply

      • Hi Charles, thanks for your quick response! The problem with ANOVA is that in a number of studies we end up having 1 group in which all vessels are open with an arbitrary value of 1900 seconds. This to me does not fit the “Within each sample, the observations are sampled randomly and independently of each other”, because the value is not really random but assigned by me. It could be that next year we decide to go for 2400 as our limit so is arbitrary. It is like if I have all groups good for ANOVA except this one in which I arbitrarily assign a value. My concern, in any case, is not if that group is really different from the control (I mean I decide those values), but if because of the arbitrary group that I have put it messes up all the follow-up tests. I hope I`m making sense. I mean in the most extreme case I can have a group of control with its own variance, and then a treated group in which all values equals 1900 seconds because all vessel stays open indefinitely. How do I analyze a situation like this? I mean they are clearly different, but How do I formalize it?

        Reply

        • Alessandro,
          I am unable to recommend an analysis approach, since I don’t really understand the specific details of the situation you are describing.
          Charles

          Reply

          • OK, I want to try maybe with a different example. Let`say you are comparing the ability of some defibrillators to restart heartbeat. You start with no beat and one instrument restarts the beat in the patient after 3 mins, one after 5 mins, one after 8, but one bad instrument never restarts the beat, and the people in that group die (hypothetically). In a way, Maybe a Kaplan-Meier? Thanks Anyway

  21. Occasionally, I’ve wanted to run an A/B split test, to decide whether a new method/sales page/whatever is superior to the current method. However, I’ve never seen a way to specify the number of tests needed to reach a decision at a certain level of confidence.

    What type of test would you suggest? Do you have any examples that I could read?

    Thanks!

    Reply

    • Hi Charley,

      It all depends on which test you use. If you use a two-sample t-test as your A/B split test, then typically the sample size (assuming this is what you mean by “number of tests”) required is based on statistical power (as well as significance level and effect size); how to calculate the sample size in this case is described at
      Power of t Tests
      Sample size requirements for t-tests

      If, instead, you base the sample size on achieving a certain confidence interval, then you can use the fact the confidence interval can be expressed as sample mean +/- s.e. * t-crit where s.e. = standard deviation divided by the square root of the sample size n; t-crit depends on the significance level (i.e. one minus the level of confidence) but also on n since the degrees of freedom for the t-test is n-2. Thus, you must solve for n, which requires using numerical means (since there is not an analytic formula). How to do this is shown in the case of the one-sample t-test at Sample Size based on the Confidence Interval

      If you use a different A/B test, then the calculations will differ. E.g. for the Mann-Whitney test, see
      Power of Mann-Whitney Test

      Please let me know whether or not I am addressing the question that you are intending to ask.

      Charles

      Reply

  23. On Mac, the StdRes function is working fine, but I can’t seem to get the AdjRes to work.
    Thanks for your wonderful site and AddIn

    Reply

    • Hello Nathan,
      Glad that you like the website and add-in.
      I just checked the AdjRes function on my Mac and it seems to be working fine. What sort of problem are you having? Are you getting an error message or are you seeing error values?
      Charles

      Reply

  24. Estimado Charles,
    He instalado el paquete en office 365, siguiendo las instrucciones para el caso, pero, al tratar de utilizar la herramienta de RealStatistics, éste no arranca y debo reinstalar cada vez que cierro un libro de excel en ejecución.

    Saludos

    Reply

  25. Dear Charles
    I have forest plant coverage data by percentage in several years. How to calculate diversity indexes? I have excel data. Could you help me?

    Reply

  27. can’t wait to do logistic regression … I am not a statistician but was put in this role a year ago with the daunting challenge of learning SPSS. I got the basics done in SPSS and learned how to manipulate the software a bit, but am now in the process of eliminating SPSS. 95% of the work formerly done in SPSS could be done in Business Objects (or Excel) using the very basics of data analysis. The other 5% is a LOGISTIC REGRESSION test that I am working on, to replace SPSS solution. Looks like I found my moving buddy!

    Reply

  28. Hi Charles, is it necessary to have a software access key?
    If so, could you tell me how to get it. I am a professor of statistical hydrology at the Autonomous University of Chiapas, Mexico, and for us it is important to use your software to calculate droughts and extreme events.
    Regards

    Reply

  31. Doc Charles, good night, please how can I make a comparison of two proportions, and one of more proportions?
    For example, for two proportions in the sex variable, the proportion of men vs. women?
    For three characteristics, the ratio of the first vs the second and the third at the same time.

    Reply

  33. Dear Dr. Zaiontz,

    I have a data set with 1 dependent variable and 3 independent variables.
    All the variables are continuous.
    The 3 independent variables are metrics of a surgery. The dependent variable is the patient’s pain score.
    I ran a regression analysis, but I know that my data is not normal.
    I suppose I could transform the data, but I was wondering if there was some sort of non-parametric procedure that I could use to run such an analysis.
    My data is not from a randomized designed experiment.
    Any suggestions?
    I appreciate your insights.

    Best regards,

    Crispus

    Reply

    • Hello Crispus,
      To use regression your data does not need to be normal. The normality requirement only applies to the residuals (that the difference between the actual and predicted values for the dependent variable). Even when the residuals are not normally distributed, the predictions can be used; the usual inferences about the coefficients and confidence intervals for the predictions will depend on the normality of the residuals.
      Charles

      Reply

  34. I can’t tell you how excited I am about finding your website! I have thought about getting back into using my graduate statistics and educational measurement knowledge to learn more from the large amount of data I accumulate in my job as a school psychologist. Sadly, I’m quite rusty, so let me ask for your guidance: to begin, I am thinking I could learn something about whether there are IQ subtests which are more or less predictive of word reading ability. Is this a multiple ANCOVA? Can you tell me where to start? Many thanks!

    Reply

    • Hello Laurie,
      Glad that you found the Real Statistics website. There is a lot of information that I hope you find useful.
      Whether IQ subtests are more or less predictive of word reading ability seems like a use for regression, but I would need more information to give better advice.
      Charles

      Reply

  35. Hey Charles

    Thanks for providing such a valuable resource.

    I’ve got a data set containing forest biomass values (n = 223) across 4 separate vegetation types. The sample sizes for each vegetation type are unequal. I’m looking to see if there is statistical separation between the mean biomass values for each of the vegetation types. I’ve performed a one way ANOVA, and want to use a Scheffe post-hoc, multiple comparison test to see where the significant differences (if any) occur. The output of the Scheffe test requires the input of ‘Scheffe contrast values’. I’m a little confused at how these are calculated. Thanks

    Reply

    • Hello Michael,
      Glad that you are getting value from the resource.
      If you are only interested in multiple pairwise comparisons, then Scheffe is not the best test to use. If the homogeneity of variances assumption is met then you should probably use Tukey-Kramer (in which case you don’t need to set contrast coefficients). If the homogeneity of variances assumption i not met then Games-Howell is usually a good choice.
      Charles

      Reply

  37. Hello dear Charles,
    My research is about comparing productivity for coastal and inland wetlands, i have two case study areas, and 37 samples from the two sites (17,19), repeated for two seasons summer and winter of year 2019, now i have (stand no, season, area, phytomass of each stand (productivity)), what kind of statistics i can do to analyze my data. when i tried one way anova it gave me no signifcant differences, but in reallity i think there should be significant differences, what should i do?

    Reply

    • The test to use depends on the hypothesis you want to test. What hypotheses do you want to test? Since you are likely to be testing multiple dependent variables you may need to use MANOVA, but again this depends on the hypotheses you want to test).
      There are two possibilities: (1) you really don’t have any significant results and should accept this (and there can be many reasons for this) or (2) you are using the wrong test.
      Charles

      Reply

  38. What formula would I use for the following?

    The director of the Office of Human Development wants to know the proportion of welfare recipients who own cars. She wants to know the proportion within 3% and wants to be 95% certain. What sample size does she need?

    The sheriff takes a sample of 100 former inmates who went through a retraining program. These inmates are traced through the county’s data system to find out whether they had a job within a year. The sheriff considers anyone with a job to be a success. The sample reveals 68 of the 100 inmates have jobs after within a year.
    What is the standard error of the proportion?

    Reply

    • Ans. to first paragraph::: n=1067
      Formula: n=1.96^2*p*q/E^2;;;E=0.03, p=q=0.5(if not given)
      Next Paragraph
      p=0.68, q=1-p, n=100, z(alpha/2)=1.96 (if not given), use above formula for computing E.
      E=0.914

      Reply

  40. How do I determine the number of raters or judges to rate my 52 identified variables. Any standard required in the choice of numbers of judges to rank the variables

    Reply

  41. Charles,

    I’m having difficulty with the most recent version of Real Statistics. After deleting the prior version and installing the most recent, I have to restart Real Statistics each time I restart Excel. I need to go to Files->Options->Add-ins and uncheck the Real Statistics box; then go back through Files->Options->Add-ins and check the box. Do you have any idea what the difficulty might be? Thanks.

    Brian

    Reply

    • Hi Brian,
      Do you need to go through this procedure in order (1) to get Real Statistics to show up on the ribbon or (2) to get Ctrl-m to work. If (1), then try using the key sequence Ctrl-Shft-m instead.
      Charles

      Reply

      • Hello Charles,

        I’m having the same issue. I have to uncheck then recheck the Real Statistics box to get it to show in the ‘Add-ins’ tab and to get ctrl-m and shift-ctrl-m to work. I have to do this each time I open a new spreadsheet. I’m running Windows 10 Pro 64-bit and Excel for Office 365 MSO (16.0.11328.20418) 32-bit.

        In any case, Thank You for this great add-in.
        Ron

        Reply

  42. Charles,

    I hope all is well 🙂

    I wish to compare quantitative exam scores for two groups. Group 1 and Group 2 are pre-calculus courses that are taught with a common syllabus and have a common final exam with a grading rubric. The only difference is that both groups used different teaching resources. Group 1 (N=116) used the traditional costly textbook and homework software, while Group 2 (N=21) used Open Education Resources – OER). I want to also mention that GR1 data is a total of 4 courses / 4 different instructors. GR2 data is from just one instructor who was exploring the use of OER.

    Both GR1 and GR2 data are approximately normally distributed so a T test was done. The results from the T test (P>.05) indicated that there is no significant difference letting me know that both groups can attain the same level of achievement on the final.

    While those results do mimic what research is saying about the comparison of these two settings, I wanted to also see if there would be a difference found in the settings if I compared each course in GR1 to GR2. That is comparing GR1/Instructor 1 exam scores to GR2, GR1/Instructor 2 exam scores to GR2, GR1/Instructor 3 exam scores to GR2 and GR1/Instructor 4 exam scores to GR2.

    It turns out that 3 of the exam distributions in GR1 were approximately normal while one not normal. I ran a t-test for the 3 normal settings compared with GR2. The results indicated that two of the settings found no difference (P>.05), and one found a significant difference (P<.05). As for the one course that is not normal (N=27), which test should I use to compare the settings with GR2 (n=21 – approx. normal data)? I was thinking to use the Mann U Whitley, however, I am not sure if the distributions have to be the same (normal/normal OR not-normal/not normal). I need your help with this.

    I also want to know if it’s OK for me just to go with only reporting in my study results the outcome of the first T test I ran with GR1 (N=116) vs GR2(N=21) that found no difference in the settings,..and not reporting the last results I found when I compared each individual course separately to GR2.

    Thanks in advance for your help!

    Reply

    • Hi Tonya,
      1. Since you are performing 4 follow-up tests, you need to take family-wise error into account. The most common approach is to use a Bonferroni correction and use alpha = .05/4 = .0125 as the significance level. There are other choices. See the following webpage:
      Dealing with Family-wise Error
      2. Mann-Whitney is the appropriate test to use for the one case where the data is not normally distributed. If the data is reasonably symmetric you can probably even get away with using the t test
      3. In my opinion you should report both results.
      Charles

      Reply

  45. Hello Charles,
    First of all, thanks for this great tool. I eager to use it on my clases. I am looking for your support on how can use excel program o Real Statistic to weight data (for expample the ones that came with weight on the same variables). I have done this on SPSS with de comand weight by “factor” but dont´know how to do it on Excel. Many thanks.

    Reply

  46. Neat tool when it works. It frequently disappears from the menu bar. And the latest sweet trick, even though I have the file in the correct roaming directory, solver is checked, etc..it refuses to appear in the menu bar at all now.

    Reply

    • You can always use Ctrl-m to access the data analysis tools.
      Are you saying that using the key sequence Ctrl-Shft-m doesn’t cause RealStats to appear on the Add-ins ribbon?
      Charles

      Reply

  47. I am just a beginner for using excel and statistics. But what I have seen on this website so far makes me very happy. Everything is in a way so structured that i can start immediately with the topics. I’m interested why you developed this website. Is it from an philanthropic attitude or is there an invisible sponsor ? I am very thankful for all this work you have done…..

    Reply

    • Hello Eric,
      I am very pleased that you like the website.
      As I recount at https://real-statistics.com/appendix/author/, I started the website so that I could perform statistical analyses for my wife’s research. I continued to develop and expand the website based on two motivations: to give something back to society and to continue to learn and interact with people around the world. I have had a lot of fun doing this and have found the activity to be quite rewarding. There is no invisible sponsor.
      Charles

      Reply

      • Hi Charles,

        First of all, kudos to you! This is a very helpful tool especially for beginners.

        I tried downloading the file but I can’t seem to install it after multiple attempts. The RealStat doesn’t appear in the AddIns.

        What do I do?

        Thanks!

        Reply

        • Hello Liza,
          What do you see when you enter the formula =VER() in any cell?
          What do you see when you press the key sequence Ctrl-m ?
          Are you using the Windows or Mac version of Excel? Which Excel are you using (Excel 2010, 2016, 365, etc.)?
          Charles

          Reply

  48. I don’t see logistic regression in the tool. I’m a mac user and I opened the tool using ctrl+M. I don’t see it in the regression tab also. Can anyone help me identify? I need help in Linear probability model as well as Logistic regression model.

    Reply

  49. Hello Charles. I would like to know if you have already created a function that calculates the cummulative probability of a bivariate normal distribution.
    In the standard Excel (without the Real Stats Add in), it is already possible to calculate the cummulative normal distribution by using the following function:
    =dist.norm.n()

    I would really like to use a similar function for the bivariate normal distribution.

    Thank you very much for the Real Stats. It has helped me a lot already. In the future, I would like to provide open access solutions as well, if I have the opportunity.

    Reply

  50. Hi Charles,

    Thanks for all the great information. I have a question. I have N attributes, such as N different features of a product and I have two groups of experts who rank these attributes from 1 to N. Also, the number of experts in each group is different. For example, I have 10 in G1 and 15 in G2. So, each one rank the attributes and then I got mean for each attribute for the experts in each group. So, now I have two different ranking (that is in fact the mean of ranking for experts in each group) for these attributes.
    Which test I can use to tell me if their ranking is different? In other words, I want to know if these two groups of experts have a significant opinion about these attributes.
    Can I use Excel to perform that test?

    Thanks much in advance for your guide.

    Reply

    • Hello Sara,
      If I understand correctly, after taking the means of the rankings, you will have N mean rankings in group G1 and N mean rankings in group G2. This means that you are comparing two vectors each with N elements. This means that you essentially have two samples of size 1. I don’t know of a test which will compare samples of size one.
      Charles

      Reply

Leave a Comment