SAMPLE SIZE CALCULATION // Cálculo de tamaño de la muestra

If you need a tool to calculate your sample size, you may want to try these links:

http://www.pwpamplona.com/wen/calcu/calculadora1.htm
An easy to use calculator from Navarra (Spain). In Spanish.

http://www.stat.ubc.ca/~rollin/stats/ssize/
A series of Java Script tools by Rollin Brant (University of British Columbia) to calculate sample size in the most frequent situations.

http://www.epibiostat.ucsf.edu/biostat/sampsize.html
A very complete web link from UCSF

http://biostat.mc.vanderbilt.edu/wiki/Main/PowerSampleSize
A free program from Vanderbilt University, by William Dupont and Walter Plummer.

http://hedwig.mgh.harvard.edu/sample_size/size.html
By David Schoenfeld , from the Massachusetts General Hospital- Harvard University.

http://www.sealedenvelope.com/power_binary.php
They also have a tool to create a randomisation list:
https://www.sealedenvelope.com/freerandomiser/v1/lists



More links in Spanish:
http://www.fisterra.com/mbe/investiga/9muestras/9muestras2.asp

http://www.psico.uniovi.es/Dpto_Psicologia/metodos/tutor.7/p3.html

http://www.seh-lelha.org/tamuestra.htm



Thanks for reading!

 ----------------------------------------------------------------------------------------------------

Si quieres calcular el tamaño de tu muestra, quizá te interese probar los siguientes enlaces:


http://www.pwpamplona.com/wen/calcu/calculadora1.htm
Una calculadora fácil de usar de la Universidad de Navarra (España)

http://www.stat.ubc.ca/~rollin/stats/ssize/
Una serie de herramientas Java por Rollin Brant (Universidad de British Columbia), para calcular el tamaño de la muestra en las situaciones más frecuentes

http://www.epibiostat.ucsf.edu/biostat/sampsize.html
Un enlace muy completo de la Universidad de California San Francisco

http://biostat.mc.vanderbilt.edu/wiki/Main/PowerSampleSize
Un programa gratuito de la Vanderbilt University, por William Dupont y Walter Plummer.

http://hedwig.mgh.harvard.edu/sample_size/size.html
Por David Schoenfeld , del Massachusetts General Hospital- Harvard University.

http://www.sealedenvelope.com/power_binary.php
También tienen una herramienta para aleatorización en este enlace:
https://www.sealedenvelope.com/freerandomiser/v1/lists


Más enlaces en español:
http://www.fisterra.com/mbe/investiga/9muestras/9muestras2.asp

http://www.psico.uniovi.es/Dpto_Psicologia/metodos/tutor.7/p3.html

http://www.seh-lelha.org/tamuestra.htm


Gracias por leer!

How many tails do you have, if you’ll forgive the expression?

Q: I have been told that my distribution is “single tailed”. What does it mean? Is this something I should feel guilty about?

A: Not at all. In fact it has nothing to do with your anatomy. Here is the explanation of its meaning:
Generally speaking, when you are testing a hypothesis you want to know if there is any difference between two treatments.
If there is no difference, we call it “null hypothesis”, (“nullus” meaning “not any” in Latin), and statisticians always make this a priori assumption. If, after the statistical test, they find no difference between treatments, they assume the NULL HYPOTHESIS. If, on the contrary they find a difference, they REJECT the null hypothesis, and therefore accept the ALTERNATIVE HYPOTHESIS. That is how it works. 

Now, pay attention: If you are comparing A against PLACEBO, there is only ONE possible alternative hypothesis (The null hypothesis is A is equal to placebo, i.e. there is no difference between them):
A is better than placebo
(…but you would not even consider the possibility that placebo is better than A, because that would mean that A, instead of a treatment, is a risk factor).


But if you are comparing drug A against drug B, then we have TWO possible alternative hypotheses:
A is better than B
B is better than B

In the first case (A against placebo) we call it a SINGLE-TAILED distribution. In the second (A against B) we call it DOUBLE-TAILED a distribution.

And, to answer a message from a male reader of the blog (in case some others share the same doubt): no, double tailed distributions will not enhance your sexual life. You have a terrible misconception. 

Thanks for reading

CTSPEDIA

This is a link kindly suggested by Prof. Frank Harrel (University of Vanderbilt, USA).
CTSPEDIA is a collection of tools, educational materials and others. It deals with clinical, statistical and translational knowledge, and it is a good homepage for your internet browser too (I mean it), because it has so many useful things just a click away. CTSPEDIA is funded by an award from the NIH. Hope you enjoy it.

http://www.ctspedia.org/do/view/CTSpedia

How to learn literature searching strategies

If you feel you need an introduction to the secrets of Pubmed this is a link you will like:

http://guides.library.ucsf.edu/introtopubmed

It is a tutorial by the UCSF library. If you want to go deeper, they also have a wide array of tools to learn literature searching strategies:

http://www.library.ucsf.edu/help/tutorials

Free-to-use animations for academic teaching

http://www.wellcome.ac.uk/Education-resources/Teaching-and-education/Animations/index.htm

This is a resource from the Wellcome trust. Animations are available for DNA, Parasitic Diseases (protozoans and worms), Bacterial Diseases and Viral Diseases.

A tutorial for SPSS // Tutorial para SPSS

This link:

http://calcnet.mth.cmich.edu/org/spss/toc.htm

is a tutorial for SPSS from the Central Michigan University


This is another one from Harvard:
http://hmdc.harvard.edu/projects/SPSS_Tutorial/spsstut.shtml



Links in Spanish:

A PDF guide to SPSS 19, from Universidad de León (Spain)
http://www.unileon.es/ficheros/servicios/informatica/spss/spanish/IBM-SPSS_guia_breve.pdf

A very quick guide to SPSS 17 (just two pages long) from Universidad de Granada (Spain)
http://webcim.ugr.es/descargas/Manual_spss/breve_manual_SPSS.pdf


----------------------------------------------------------------------------------------------------
Este link

http://calcnet.mth.cmich.edu/org/spss/toc.htm

es un tutorial para SPSS de la Central Michigan University


Este otro es de la Universidad de Harvard:
http://hmdc.harvard.edu/projects/SPSS_Tutorial/spsstut.shtml


Enlaces en Español:

Una guía en PDF para SPSS 19, de la Universidad de León (España)
http://www.unileon.es/ficheros/servicios/informatica/spss/spanish/IBM-SPSS_guia_breve.pdf

Una guía muy rápida (sólo dos páginas) para SPSS 17 de la Universidad de Granada (España)
http://webcim.ugr.es/descargas/Manual_spss/breve_manual_SPSS.pdf

COMMITTEE ON PUBLICATION ETHICS

http://publicationethics.org/

Information on most aspects of publication ethics. It includes guidelines, flowchart, e-learning and code of conduct.

Using PRISMA for systematic review reporting

http://www.youtube.com/watch?v=TVFYenon1Jo

A Youtube link to a tutorial of about one hour about the use of the EQUATOR network tool called PRISMA for reporting systematic reviews. By Prof. Moher from the University of Ottawa.

NOBODY IS PERFECT

Nobody is perfect

Or, in other words, we all make errors. Let’s see the types we can make:

Suppose you are testing the effectiveness of drug A compared to drug B. There are two real possibilities: either there is a difference between them or not.
After you perform the test, there are two possibilities: either you FIND a difference, or not.
You could put that in a 2x2 table:

                                   Real difference    No real difference
Computer says YES      Well done!         Type I error
Computer says NO      Type II error      Well done!

There are two ways you can go wrong. If you affirm there is a difference between A and B, and there is not, you have a type I or alpha error. If you are not able to find a real difference between A and B, you have a type II or beta error.

You can put it in other words:
If you reject a null hypothesis but you should have accepted it, you fall into a type I or alpha error. If you accept a null hypothesis but you should have rejected it, that’s a type II or beta error.

There is a mnemotechnic for it; Beta stands for Blind. (Beta error means you are blind to see a real difference between the two drugs).


Thanks for reading!

Free software form CDC. And more...

The Center for Disease Control and Prevention (CDC) offers a well known free statistical program, called Epiinfo. You can access it here:

http://wwwn.cdc.gov/epiinfo/

More resources:

Openepi is a free software program developed by the Rollins School of Public Health from Emory University in Atlanta, GA.

http://www.openepi.com/OE2.3/Menu/OpenEpiMenu.htm

Epidata. Free software by JM Lauritsen et al., from Denmark.

http://www.epidata.dk/

Netepi
http://sourceforge.net/projects/netepi/





Software for Systematic reviews and meta-analysis:

Revman is the software from the Cochrane Collaboration to feed in data and produce a systematic review:
http://ims.cochrane.org/revman


Epimeta, from the CDC
http://ftp.cdc.gov/pub/Software/epimeta/


MIX
http://www.mix-for-meta-analysis.info/

Software for Clinical Trials:

https://community.openclinica.com/

Healthknowledge and more...

Try these: 

http://www.healthknowledge.org.uk/e-learning/statistical-methods
A fully furnished, higly reliable and useful website. Free online course on statistical methods for Health Care workers. Tutorial on Evidence Based Medicine.

http://www.cebm.net/
Centre for Evidence Based Medicine of the University of Oxford.

http://www.gradeworkinggroup.org/
GRADE wepage. GRADE is a system to rank evidence and recommendations for a given heath problem. It is used by most major medical journal and evidence based medicine organisation.

http://www.epibiostat.ucsf.edu/biostat/sampsize.html
The website of the Division of Biostatistics of the University of California, San Francisco

http://www.ucm.es/info/matbio/piembb/enlaces/Free%20Statistical%20Software.htm
A site from the Universidad Complutende de Madrid, with software links.


http://www.stat.ufl.edu/vlib/statistics.html
This is a vast list of links to many statistical resources, from academic and government departments to software or stat journals. Published by the University of Florida.

What is an epidemiologist?

A clear definition of what an epidemiologist is from a lecture at Columbia University! 2 minutes long.

http://www.youtube.com/watch?v=jCK2mflwESM




By the way, the link to Mailman School of Public Health at Columbia University:

http://www.mailman.columbia.edu/academic-departments/epidemiology

Books: free links

Free links to Epidemiology and Statistics books:

http://publichealth.massey.ac.nz/publications/introepi.pdf

Neil Pearce´s “A Short introduction to Epidemiology”

http://faculty.vassar.edu/lowry/webtext.html

By Richard Lowry, from Vassar College

http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=12156

“Street-fight Mathematics”,

by Sanjoy Mahajan, from MIT

http://www.math.ku.dk/~sjo/papers/HaldBook.pdf

“A History of Parametric Statistical Inference from Bernouilli to Fisher, 1713 – 1935”. By Anders Hald

Graphics and data display

Here goes a link to the latest version of a pdf book on statistical data display, from Rafe Donahue (Vanderbilt University Medical Center).
with tips and advise for those who want to show their study results in a clear elegant manner.
http://biostat.mc.vanderbilt.edu/twiki/pub/Main/RafeDonahue/fscipdpfcbg_currentversion.pdf


Also check this link to Tufte's book:
http://www.amazon.com/Visual-Display-Quantitative-Information/dp/0961392142

International statements

These are international statements for the reporting of different types of studies.

Diagnostic studies:
http://www.stard-statement.org/

Systematic reviews:
http://www.prisma-statement.org/

Observational studies:
http://www.strobe-statement.org/

Clinical trials:
http://www.consort-statement.org/

We need more biostatistical training. And are aware of it

This paper from the Mayo Clínic Proceedings shows that mosto doctors feel they should improve their biostatistical training. They also find that Statistics is an important and useful tool for their careers.

http://www.mayoclinicproceedings.com/content/82/8/939.full

P for novices // P para novatos

p...
What is p value (also called significance)? How does it work? Curiously enough, many collegues at the hospital clearly know how to interpret a p value (i.e, they can tell you if a study is significant or not), but have trouble trying to explain the exact meaning of, say, a p = 0. 01.
In practice most of us would answer: if p is lower than 0.05, the result of the study is significant. If higher, non significant. Simple. But what does it mean?
So let´s go for it. Imagine we are comparing two treatments A and B. And imagine there is no real difference between A and B, in terms of their efficacy. We call p the probability that,  when you perform a statistical test (whichever you need, we will try to roughly cover that in a future post), you FIND a difference by chance.
In other words, when you perform your test you have a chance that it does not show reality, but on the contrary get a false result. And, as there is no difference between A and B, that false result means that you SEE a difference.
And p value has nothing to do with the difference of the effect found (A being 10%, 30% or 70% more effective than B). You may find a tiny difference between the effectiveness of the two drugs, with a very significant p value. Or a huge difference with a clearly non significant p.
In fact p is like a friend. You can trust or not your friend (that is the p value) on what he tells you (that is the difference between A and B that you find; 2%, 25%, or whatever).
Why 0.05? Well that is a value everyone has agreed upon. It does not come from a formula or anything. When we try our test, if p= 0.05, it means that we want to be wrong once every 20 times at most.


Was this post useful to you? Would you like to talk about any specific subject? Please share your comments!

-------------------------------------------------------------------------------------------------------------------------------
p...
¿Qué es la p (también llamada significación estadística)? ¿Cómo funciona? Cusiosamente, muchos colegas del hospital saben interpretar un valor de p (por ejemplo, pueden decir si un estudio es significativo o no) pero tienen problemas para definir el significado exacto de una p de 0,01 (por ejemplo).
En la práctica, diríamos que una p menor de 0,05 es significativa, y mayor, no significativa, pero ¿que quiere decir esto?
Vamos allá. Imaginemos que tenemos dos tratamientos A y B, y que no hay diferencias en cuanto a su eficacia (son igualmente eficaces). Llamamos p a la probabilidad de encontrar una diferencia entre A y B simplemente por azar (recordemos que no hay diferencias entre A y B).
El valor de la p no tiene nada que ver con la diferencia entre tratamientos (entre A y B puede haber una diferencia de eficacia del 10 , del 30 o del 70%). Por ejemplo, se puede encontrar una diferencia muy pequeña entre A y B con una p enrormemente significativa, o una gran diferencia con una p claramente no significativa.
En realidad p es como un amigo. Lo que nos dice nuestro amigo es la diferencia entre A y B. La probabilidad de que se equivoque es la p.
¿Por qué 0,05? Por puro acuerdo. No viene de de ninguna fórmula. Cuando p= 0,05 significa que en nuestro test estadístico nos vamos a equivocar 1 de cada 20 veces.

Free (and not so free) online resources

This is a review of online biostatistics links and resources. Some of them are free; some others are not, although they are worth the money.

http://ocw.jhsph.edu/courses/introbiostats/

An Opencourseware free course offered by the Johns Hopkins School of Public Health. They also offer two courses on Methods in Biostatistics:

http://ocw.jhsph.edu/courses/MethodsInBiostatisticsI/

and fundamentals of epidemiology:

http://ocw.jhsph.edu/courses/FundEpi/

http://ocw.mit.edu/courses/mathematics/18-05-introduction-to-probability-and-statistics-spring-2005/

The OpenCourseWare (OCW) is an initiative started by the prestigious Massachusetts Institute of Technology. It is a free collection of courses previously lectured at the MIT. A great idea followed by other universities. MIT OCW offers a broad array of courses on many disciplines, and they have this course titled “Introduction to probability and Statistics”. Free of charge.

http://ocw.tufts.edu/Course/1/Coursehome

Another OCW resource, this time from Tufts University. Highly recommendable.

http://isites.harvard.edu/icb/icb.do?keyword=k83983

An online course from Harvard, titled Elements of Clinical Investigation. Charge ranges from $1,025 (non-credit) to $1950 (if you are aiming for graduate credit).

http://isites.harvard.edu/icb/icb.do?keyword=k84313

An introductory course on Statistics from Harvard. Same price as above.

http://extension.berkeley.edu/catalog/course26.html

A 785$ course offered by the University of California Berkeley. You can enroll anytime, and have 6 months to finish it.

http://bibalex.org/supercourse/courses/biostatistics1.htm

A bilingual course (in English and Spanish) by Dr. Nicolas Padilla Raygoza, a Mexican pediatrician and epidemiologist. Free.

http://oli.web.cmu.edu/openlearning/forstudents/freecourses/statistics

Two free online courses from the Carnegie Mellon University. One is on Probability and Statistics and the other on Statistical Reasoning. You need to log in to take them, but at no charge.

New Year Address

This new blog seeks to share information about medical statistics, evidence based medicine and research methodology. Everyone is welcome to share and comment. All constructive criticism is encouraged!

On one hand anyone can upload (and find) resources, tools, publications of a certain "academic" flavour.
On the other hand this intends to be a forum to publicly comment any health subject where a methodology issue should be improved.

And as the first post of the year, I would like to share the outstanding blog by Ben Goldacre, BAD SCIENCE:

http://www.badscience.net/




Cheers