Welcome to all statisticians who study in area of theory or application.

• Muhammed Sollom

  How can we deal with highly scattered data when the mean in this case is smaller than the variance?

  Data analysis

  Recent replies ⋅ Show All (0)

  • 
    Emmanuel Curis replied

    Well, without more precision about the problem, I would answer that I see no theoretical concern with having higher variance than mean, especially knowing they are not in the same unit so difficultWell, without more precision about the problem, I would answer that I see no theoretical concern with having higher variance than mean, especially knowing they are not in the same unit so difficult to compare... Neither with having higher standard deviation than mean, by the way.

    2 hours ago
• Gopinath Panda

  Which is a good book on M/G/-/-/ queueing models for beginners?

  Recent replies ⋅ Show All (4)

  • 
    Gopinath Panda replied

    Sir i have read these models from Gross and Harris book. But i couldn't understand the alalysis of M/G/-/- model.

    4 days ago

• Post protected

  Join ResearchGate now to read this post.

  Recent replies ⋅ Show All (8)

  • 

    Comment protected

    Join ResearchGate now to read this comment.
• Francisco Cuevas

  How can i simulate Outliers in a regression model?

  I am testing the M-estimators for the linear regression, but I don't know how to generate outliers.

  Recent replies ⋅ Show All (10)

  • 

    Comment protected

    Join ResearchGate now to read this comment.
• Jason Leung

  Would I use independent 2 samples or paired samples T test to test for mean difference between age-sex-matched samples with ratio 1:2?

  I have a case group and a control group, which is age-sex-matched with ratio 1 case: 2 control. I'm wondering what test is suitable to test their mean difference. To test matched groups, some suggestI have a case group and a control group, which is age-sex-matched with ratio 1 case: 2 control. I'm wondering what test is suitable to test their mean difference. To test matched groups, some suggest paired t test, some use independent t test. But I find difficult to use paired t test because they are not 1:1 but 1:2. I found it is difficult to treat it as matched sense if it is not 1:1 ratio. Any idea? Thanks

  Recent replies ⋅ Show All (41)

  • 
    Mikhail Nikulin replied

    I think that the T-test is used for the normal samples.

    12 days ago
• Sunil Kumar

  How do you determine a sample size for an infinite population?

  How do you determine a sample size for an unknown population that can state with a predetermined confidence?

  Recent replies ⋅ Show All (4)

  • 
    Keith D Baldwin replied

    Also check out this reference. It is a systematic review which suggests that one half a sd may be appropriate for judging clinical significance. This correlates with a cohens d ofAlso check out this reference. It is a systematic review which suggests that one half a sd may be appropriate for judging clinical significance. This correlates with a cohens d of 0.5. Interpretation of changes in health-related quality of life: the remarkable universality of half a standard deviation. Norman GR, et al. Med Care. 2003  Page 1 of 1 

    14 days ago
• Judit Camacho

  Where can I obtain non-parametric goodness-of-fit tests written in fortran codes?

  Such as Cramér-von Mises test, K-S test, etc. I have my observational data. I have done some Monte-Carlo simulations (which take long time) with different initial conditions and then have ISuch as Cramér-von Mises test, K-S test, etc. I have my observational data. I have done some Monte-Carlo simulations (which take long time) with different initial conditions and then have I bootstrap them. So I get possible ways to reproduce my observational data. I have to decide which ones are the highly likely options. I am using KS-test but it would be nice if different nonparametric tests were applied. Which others do you recommend? Which other tools may be useful?

  Recent replies ⋅ Show All (6)

  • 
    Charles Warner replied

    Sponsored by the United States National Institute of Standards and Technology, the site, NIST Digital Library of Mathematical Functions (http://dlmf.nist.gov/), is a digitized version of the standardSponsored by the United States National Institute of Standards and Technology, the site, NIST Digital Library of Mathematical Functions (http://dlmf.nist.gov/), is a digitized version of the standard mathematical reference, the Handbook of Mathematical Functions with Formulas Graphs, and Mathematical Tables, edited by Milton Abramowitz and Irene A. Stegun. The web site includes a link to the Guide to Available Mathematical Software (http://gams.nist.gov/), which is a cross index of mathematical software in use at NIST, and includes many Open Source implementations of the procedures detailed in the Handbook, many written in Fortran. Another possible source you might consider would be http://netlib.org/.

    27 days ago
• Miguel Pessanha Pais

  Is there a way to do cluster analysis while forcing replicates to cluster together?

  I have 6 measures per case, with 10 cases and I want to run a non hierarchical cluster analysis such as k-means (preceded by an analysis to define the number of clusters k that minimizes within-groupI have 6 measures per case, with 10 cases and I want to run a non hierarchical cluster analysis such as k-means (preceded by an analysis to define the number of clusters k that minimizes within-group sum of squares) on the 60 values, but somewhat forcing cases to stick together, so that for a single case I don't get 3 values going to one cluster and 3 values to another... In some cases I get a majority (4 or 5 out of 6) of values clustering together, but I don't know if it is even ethical to assume that this case belongs to cluster x because the majority of values from that case went to cluster x... So in other words, I wish to cluster cases, while taking into account their inherent variation, represented by the 6 values. Is there a way of doing this? Or a different method?

  Recent replies ⋅ Show All (12)

  • 
    Miguel Pessanha Pais replied

    Yes, I sent you these examples because they show the kind of data I am facing (attributes with many zeros, outliers). In fact that replicate is an outlier in lots of attributes, and I know why.Yes, I sent you these examples because they show the kind of data I am facing (attributes with many zeros, outliers). In fact that replicate is an outlier in lots of attributes, and I know why. Excluding is the best option indeed, because I still have 5 left to "represent" the location. I have excluded it in many attempts I have done to treat my data before this, but as some methods just kick outliers out to a cluster of their own, I kept the data completely raw in those sheets. The attribute with lots of zeros can be taken out, I guess. In some distance measures it would not make a difference, in others it would perhaps bring cases a bit closer together, but excluding the attribute completely is probably the best solution. Thank you for pointing that out. :)

    28 days ago

• Post protected

  Join ResearchGate now to read this post.

  Recent replies ⋅ Show All (22)

  • 
    Yar Yot replied

    You will try to monitor movement

    28 days ago
• Divya Sebastian

  Power Analysis for Sample Size

  Hi can someone guide me, please..! Want to know how to use power analysis for fixing the sample size...what are the different criteria's that I should take into consideration?Is any softwareHi can someone guide me, please..! Want to know how to use power analysis for fixing the sample size...what are the different criteria's that I should take into consideration?Is any software available for doing Power analysis.How reliable is it?

  Recent replies ⋅ Show All (14)

  • 
    Antonin Cuc replied

    I studied at the Czech Technical University Prague in the field of innovation management in industry and thus I passed about 5 semesters of statistics, mainly from the statistical acceptance ofI studied at the Czech Technical University Prague in the field of innovation management in industry and thus I passed about 5 semesters of statistics, mainly from the statistical acceptance of goods, from the use of observational "flash methods" for studying complex processes, foundations of usage the Monte Carlo simulation. On the other hand, I tried to address a very practical problems in the machinary workshops that still no one before me did not address - for example, we could get statistically conclusions from the analysis of a random sample of archived living engineering drawings made ​​part of the engineering company, to recognize the possibilities in innovations. It was very exciting to me - and actually classically trained statisticians did not want to work too - because I was constantly trying to apply statistical methods to very extreme applications by no typical statistical tasks in garanted statistical perfect conclusions. For example - some observed statistical samplings of quantity 150; 350; 600 entities in the sample - of the archived technical life drawings - as a sufficient ground for some experimental interpretations, but some aspects were just a good guide with one sample, and for other conslusions it was only fuzzy information. Sometimes we managed to work with a point estimate of the parameter include with an attempt confidence interval exactly, but some other parametr was as inputs information s statistically weak and it could be sampling only for a introductory image about a range of value, etc - and the usage required an intuitive illustration to decision and interpretation must be realized with more creativity and uncertainty. This led me to discuss or understood - what is a good sample of whot is good - statistically manageable task, what information is vague and what is solid and what is only a speculation or whot could maybe desinterpetation. It was an amazing experience that with me but unfortunately statisticians were not willing to share - so I implemented projects for innovations in machinary workshops alone, that has been very extremaly successful in comparing with a traditional design - quicker and very cheap, in time - but now I am old man - and probably no one has ever learn my method of deciding S_T_A_R_S_Statistical Retrieval and Search of information to Decision Making, Utility Model 21532 Czech Republic. I am working with sufficient statistic Inputs of information to solve repeated strategic decisions with sufficient statistic representative conclusions and I try to used optimal Sources of information to gain Inputs with reliable information. It is a new princip to create new automatic information systems and quite a different communications in distributed databases with computer support to decision making. I can very short time to begin design work - with changes statistic reliability step by step to solve informational complex background to very complicated strategic decision with a choice the best one older or we could partly derive an excelent new decision with minimum risk to accept bad decision. We respect working Shanon´s transformal channel with minimum redundancy and increses of entropy to fullfil hierarchical targets and beeing stability in long time with respects to interactions with neighbourhood

    Apr 26, 2012
• Judit Camacho

  How many bootstraps can I perform on a sample of N elements?

  I am using bootstrapping analysis for a set of data that I obtained from a Monte-Carlo simulations. Bootstrapping (statics) allows random sampling with replacement from the original data set that II am using bootstrapping analysis for a set of data that I obtained from a Monte-Carlo simulations. Bootstrapping (statics) allows random sampling with replacement from the original data set that I obtain from a Monte-Carlo simulation. Thanks!

  Recent replies ⋅ Show All (34)

  • 
    jean-louis Foulley replied

    Sorry It is for sure 35 not 24.

    Apr 25, 2012
• Antonio Irpino

  Density estimation or probability estimation?

  I am working on histograms because histogram is a very parsimonious way of storing a distribution of observed values. In order to overcome the problem of the choice of the width of bins, I devised aI am working on histograms because histogram is a very parsimonious way of storing a distribution of observed values. In order to overcome the problem of the choice of the width of bins, I devised a method where, chosen the desired number of bins, the domain is binned into bins that have different width. OK, nothing new, just a piecewise interpolation of the distribution function. But I intend to compare the procedure against other methods. The kernel density estimation looks the more competitive (a lot of references stated its superiority in being consistent and having a fast convergence rate to the "true" density). However, I performed a test to asses the superiority of KDE vs "my histogram". I generated 1million points from a mixture of two normals. I performed the KDE with the RBF kernel storing the distribution function estimated into 500 points. I estimated "my histogram" with just 16 bins. Well, after that I simulated 10k random queries about the probability of an interval of values. Whit my great surprise, "my histograms" is more accurate (MSE) in the prediction of the probability than the KDE. So, my question is: "Is it more correct (and\or useful) to predict density or probability?" If you are curios of that, I have implemented the procedures in MATLAB.

  Recent replies ⋅ Show All (2)

  • 
    Burak Alakent replied

    I've had some experience with (Gaussian) kernel density estimation, and I can say that it is superior to histograms, UNLESS there are hard limits on the tail of the distribution. One (simple butI've had some experience with (Gaussian) kernel density estimation, and I can say that it is superior to histograms, UNLESS there are hard limits on the tail of the distribution. One (simple but informative) example might be grades from an exam, which cannot be lower that zero, and cannot be higher than 100. Histogram representation would be more realistic for such a case, whereas the curve obtained by KDE would reach negative numbers and numbers greater than 100 . Another example, which I encounter many times in my research, is the distribution of atomic distances, which cannot be less than the sum of van der Waals radii of atoms. Tails of KDE of these distances would include those "forbidden" regions, which is not realistic. I may say that KDE works better if the underlying p.d.f. is nonzero for all real numbers (or within the region of interest). Otherwise, I'd prefer histograms.

    Apr 24, 2012
• Giovanni Leopoldo Rozza

  Normal Distribution Probabilty Density Function Properties

  If I have two normal probability density functions f1(x)=N1(a1,b1) and f2(x)=N2(a2,b2). a1,a2 = mean b1,b2 = variance Can I say that a new probability function N3=(a1-a2,b1-b2) = f1(x)-f2(x) forIf I have two normal probability density functions f1(x)=N1(a1,b1) and f2(x)=N2(a2,b2). a1,a2 = mean b1,b2 = variance Can I say that a new probability function N3=(a1-a2,b1-b2) = f1(x)-f2(x) for each x ? Thank you Giovanni

  Recent replies ⋅ Show All (4)

  • 
    Antonio Irpino replied

    Please, write better the question. Theoretically if the two distributions are uncorrelated x3 is a normal whit mean a1-a2 and standard deviation sqrt(b1^2+b2^2) (in general, if they are notPlease, write better the question. Theoretically if the two distributions are uncorrelated x3 is a normal whit mean a1-a2 and standard deviation sqrt(b1^2+b2^2) (in general, if they are not uncorrelated the standard deviation is sqrt(b1^2+b2^2-2*covar(x1,x2)).

    Apr 21, 2012
• Michael Fauler

  Comparing two correlation coefficients (Kendall's Tau) from control and treatment group

  How can I perform a hypothesis test on the difference of Kendall's Tau between independent control and treatment/disease group. Can I use Fisher's Z transform as for Pearson's R (=Steiger's Z-test)How can I perform a hypothesis test on the difference of Kendall's Tau between independent control and treatment/disease group. Can I use Fisher's Z transform as for Pearson's R (=Steiger's Z-test) or do I have to bootstrap, but how?

  Recent replies ⋅ Show All (1)

  • 
    Caetano C.R. Penna replied

    I just wanted to find out the same thing and came across ResearchGate, so I joined. Hope you will get an answer. :-) Let me know if you find out about it - I'll do the same.

    Apr 21, 2012
• Mateus Ricardo Nogueira Vilanova

  Ranking the best fit of statistical distributions

  Dear colleagues: the adjustment of continuous distributions like GEV, Gumbel and Log-Pearson3 is a common practice in hydrology, for the definition of flood peak discharges. The literature of theDear colleagues: the adjustment of continuous distributions like GEV, Gumbel and Log-Pearson3 is a common practice in hydrology, for the definition of flood peak discharges. The literature of the area provides some classical tests of adhesion (chi-square, Kolmogorov-Smirnov test - that considers the greatest differences between the theoretical and empirical distributions - Anderson-Darling - which gives more weight to the tails of the distribution, mainly). My question is: considering that I made the adjustment of various distributions, and that all three tests (X2, KS and AD) concluded that all distributions are acceptable (for an alpha of 0.05 for example), how can I set what is THE BEST DISTRIBUTION? Analyze and compare the p-values ​​for each test, for each distribution, can be a way? For example, if the GEV had the higher significance according to the X2 test, Gumbel the higher significance according to the KS test and Log-Pearson3 the higher significance according AD, which would be the best distribution for my dataset? Thank you!

  Recent replies ⋅ Show All (11)

  • 
    Thomas P A Debray replied

    Irregardless of the techique you use for evaluating model fit, be aware there is an issue of multiple testing here. this implies that you should not directly interpret significant fits with betterIrregardless of the techique you use for evaluating model fit, be aware there is an issue of multiple testing here. this implies that you should not directly interpret significant fits with better "likelihood" as truely better models. An increased amount of tested models increases the probability of random findings.

    Apr 20, 2012