Michael Chernick

19,793 reputation

21539

bio	website
	location
	age
visits	member for	1 year
	seen	Nov 14 '12 at 14:43
stats	profile views	6,106

I am a biostatistician at the Lankenau Institute for Medical Research where I work on lab experiments, clinical trials and other medical research. I have a PhD in Statistics from Stanford University. I have published books on bootstrap and biostatistics and have written or coauthored many articles in statistics, mathematics and medical journals. I am an ASA Fellow and am also a member of ENAR, the IMS, the Bernoulli Society and the Royal Statistical Society. I like teaching and mentoring and playing chess with my son (who usually beats me).

	19,793 reputation	bio	website		visits	member for	1 year
21539 badges		location			seen	Nov 14 '12 at 14:43

summary answers questions tags badges favorites bounties reputation activity

4,661 Actions

suggestions reviews revisions comments badges posts accepts all

Sep 19	comment	How do I find peaks in a dataset? @whuber After giving more thought to your statement "What I'm trying to suggest, Michael, is that in this case the nonparametric approach is likely to be far better than any parametric approach except when the data cleave especially closely to the model--and even then it will perform well." I remarked that this was an opinion but I think I should object to it more strongly. During my 5 years as a mathematician at Aberdeen Proving Ground (before I could call myself a trained statistician) I had occasion to see many real world examples where exponential smoothing and ARIMA modeling were used.
Sep 19	answered	Investigating the change in a proportion/ratio
Sep 19	reviewed	Approve suggested edit on R-code question: model selection based on individual significance in regression?
Sep 19	comment	How to calculate weighted Hedges' g effect size in meta-analysis when some effect sizes share a control group? 3. Handbook, co-edited by Hedges
Sep 19	comment	How to calculate weighted Hedges' g effect size in meta-analysis when some effect sizes share a control group? I don't know the answer to your question but Larry Hedges has coauthored books on meta-analysis and it would seem those books would be good places to look to find your answers. Here are the links: 1. Olkin and Hedges 2. Borenstein, Hedges et al.
Sep 19	reviewed	Reject suggested edit on Whether to use correlation matrix between latent variables or raw data as input to SEM?
Sep 19	comment	Contradiction between significant effect in multiple regression, but non-significant t-test on its own No I am talking about the mean response for males - mean response for female which is determined not to be significant based on the t test.
Sep 19	comment	Estimate the proportion and variance in a simple binomial cluster design The sampling is just k independent sets of independent identically distributed Bernoulli trials. So it amounts to the same thing as one sample of N independent identically distributed Bernoulli trials.
Sep 19	answered	Contradiction between significant effect in multiple regression, but non-significant t-test on its own
Sep 19	answered	Estimate the proportion and variance in a simple binomial cluster design
Sep 18	answered	Parameter estimation for linear system with correlated noise
Sep 18	comment	Parameter estimation for linear system with correlated noise If the errors were correlated over time you could model the noise component as a mving average with the coefficients of the moving average determining the autocorrelation function for the noise. Model parameters for time series which include the coefficients of the moving average terms can then be fit using maximum likelihood. If the data do not form a time series there still may be a way to model the noise term so that correlation is part of the structure.
Sep 18	revised	Parameter estimation for linear system with correlated noise added 1 characters in body
Sep 18	comment	Individual patient data meta-analysis with SD's as outcome measure If s$_1$ is the standard error of mean 1 and s$_2$ is the standard error of mean 2 did you take into account that the standard error for the mean difference (given the samples are independent is √[s$_1$$^2$ +s$_2$$^2$] which is bigger than either s$_1$ or s$_2$?
Sep 18	comment	Individual patient data meta-analysis with SD's as outcome measure Do you get estimates in standard deviation units if you divide the scores by the indicated standard deviations? Effect size is generally defined as the mean difference /standard deviation of the estimated mean difference. So effect size can be viewed aas being in standard deviation units.
Sep 18	comment	Using the Fisher's sign test to select the better of two classifiers If the number of +s is about the same as the number of -s the differences could be just due to chance. If they are very difference there could be a real difference in performance. But I think a better way would be to construct a 2x2 contingency table of predictions and test for discordance in the contingency table or you could apply Cohen's Kappa test for agreement/disagreement. If you want to compare performance look for significant differences in their error rates.
Sep 18	comment	Using the Fisher's sign test to select the better of two classifiers If you are pairing the classifiers on their predictions for common cases then the sign test could be used to test that the algorithms differ. Given a score 0 for the first class and 1 for the second class you can take classifier 1s score and subtract classifiers 2s score 0 doesn't count. You just keep track of the number of -1s .
Sep 18	comment	Using the Fisher's sign test to select the better of two classifiers There are tests attributed to Fisher and there is a sign test. But I have not heard of a test referred to as Fisher's sign test.
Sep 18	comment	Individual patient data meta-analysis with SD's as outcome measure And in units of standard deviation theses scores would be 60/25=12/5=2.4, 43/25=1.72 and 30/25 =6/5=1.2.
Sep 18	comment	With simple random sampling, how to approximate variance of R=avg(Y)/avg(X)? But this will not be a problem if the Xs and Ys must all be nonnegative. Then both U and V will be bounded away from zero. We really only need this to be the case for V.