Niclas, non-statistics students (like students of psychology, life sciences etc) will actually never really need the math behind statistcs, and teaching the math will hardly improve any understanding (you may give me a counter-example where this really worked!). But here again I'd like to emphasize that teaching students how to perform tests is just a highway to "mindless statistics", to educate essentially statistically illiterates that do stats only because others do it and in the way others do it and perform tests just because reviewers like to see them - without even vaguely understanding what empirical science is about: namely about cumulating knowledge. This seems to be trivial, but understanding this requires a lot of things (what is knowledge? - knowledge is represented by models - what are models? - what is information? what is evidence? how does this relate to data? etc etc etc). THIS is what should be taught in statistics. When stats education ends up in researchers that say "I have some data here. How do I test it?" one can be sure that nothing had been learned or understood and statistics will remain a stupid tool to satiffy reviewers instead of a basis to think about experiments, data, and conclusions (about science). The results will neither be meaningful nor representative.

Hello, Michael -

Funny thing happened yesterday... My son and I were talking about how statistic courses are taught to grad students, and the problems that these students are faced with later on when it is time to write their theses or dissertations.

I am a PhD candidate in Chapter 4 of my dissertation and I am also mentoring a group of my peers through this process. None of us had previous experience with statistics beyond a stats class in our College Math course, so the concepts, terms, and methods of analysis were almost totally long forgotten, or foreign to us.

I have my own ideas on what the weaknesses are in how research classes are designed. I do not think that grad students are fully prepared for the research and analysis part of their educational journey. Is this the approach that you might be considering? For example, do you plan on offering a method (such as with a new course design) that makes more sense for students who may never have had much experience with statistics? Or is it to be a "how to" book that  teaches how to do the different types of analyses - or both? Very interested in understanding  more....

Ruth

I'm teaching statistics to undergraduate psychology students and I agree with Ruth, it is difficult especially when the students don't have a Math background in their previous education. It is even more difficult because at the beginning they are not aware of the usefulness of the research. But my experience until now showed me that it is not impossible. I try to find for each class interesting application in order to stimulate their learning motivation and their interest toward statistics. 

Interested in learning more similar to Ruth. I completed my Ph.D. 6 years ago and noticed vast differences (continue to notice in my career working with other Ph.D.) in graduate preparation across education programs.

https://www.researchgate.net/post/Do_we_actually_need_or_understand_more_than_basic_statistics_for_routine_investigation?_tpcectx=profile_questions

Hi Michael,

The above question thread may give a counter-view to how we teach statistics & what we focus on.  Unless a student is studying the field of statistics specifically, I would prefer them to know what tests do, when to use them & how to use them, before they get to the 'exiting bit' of complex mathematics. You might say that they need to know how statistics work before they can use them, but unless this is central to their studies, it is like saying you have to do complex math without a calculator or know how your computer programmes function internally.  If the focus is placed on utility, then we will hopefully generate meaningful & representative statistics  ;-)

Niclas, non-statistics students (like students of psychology, life sciences etc) will actually never really need the math behind statistcs, and teaching the math will hardly improve any understanding (you may give me a counter-example where this really worked!). But here again I'd like to emphasize that teaching students how to perform tests is just a highway to "mindless statistics", to educate essentially statistically illiterates that do stats only because others do it and in the way others do it and perform tests just because reviewers like to see them - without even vaguely understanding what empirical science is about: namely about cumulating knowledge. This seems to be trivial, but understanding this requires a lot of things (what is knowledge? - knowledge is represented by models - what are models? - what is information? what is evidence? how does this relate to data? etc etc etc). THIS is what should be taught in statistics. When stats education ends up in researchers that say "I have some data here. How do I test it?" one can be sure that nothing had been learned or understood and statistics will remain a stupid tool to satiffy reviewers instead of a basis to think about experiments, data, and conclusions (about science). The results will neither be meaningful nor representative.

Dear Ruth, and Lauren:

Just to let you know a bit more the kind of book I am thinking could have the following features:

1) Population goal: PhD students, postdocs

2) More maybe a didactical approach than a "how to book"

3) Goal: people understand the way "experts" (whoever they are, but certainly not the ones that called themselves experts) in the field thing, how they approach problems and overall, what represents a "problem" for this field. Secondly, to unify the diversity of technical language around this field, to differentiate the different disciplines that contribute to,say, "statistical thinking" if you wish - this includes the myth of math and clarifies what place of maths is in stats.

This is so far the stand of the idea.

Thanks for your interest. If you want to engage in this feel free to email me.

Michael.

Nicholas & Jochen:

You seem to have quite different views on teaching stats. The only thing that I can say is that I disagree with the idea that "we experts" decide what to teach to whom. That means, for students of statistics, they will a kind of teaching that will allow them to understand, the rest of us, just "click here and there in your spss". I do not agree with that basic approach and for someone who so thinks, needs to justify/argument why things should be so. I can not be less interested in "utility", I am interested in trying to show the beauty behind such endevour as statistical thinking - among others.

Maybe the book title could be "from mindless to mindful statistics":)

Types of texts are not as important as the approach used in teaching statistics to graduate students. It is also important here to separate the target audiences. Statistics graduate students need the theoretical background in order to be able to manage the Ph.D. course work and in order to have the knowledge to one day become Statistics Professors or researchers at some agency. 

Back to the idea of writing a new text on statistics. I would make each chapter built around some application. Encourage the use of Google to look up information on statistics and to use Google Scholar to find references.

Dear Raid,

Thanks for drop a line. However, I disagree with your separation of text versus teaching approaches. I have taken several courses at the phd level and all of them used a basic text book to organize their teaching. Moreover, I was involved in a project that tried to get school  students to reflect and get deeper in mathematics, it was very difficult and one of the causes of that was simply that the curriculum and the material was too "procedural", and not focused on "higher order thinking" so to speak. So, materials like text book do matter and to the same degree.

Have a nice day,

Michael.

Dear Michael,

I feel that it depends on the type of students and also on the level. I emphasize the choice of text for undergraduate courses and for foundation graduate courses, but when I have advanced graduate students only, then I favor a mix of articles and project based instruction.

Hi Michael, you mention some key points as:

"3) Goal: people understand the way "experts" [...] in the field think, how they approach problems and overall, what represents a "problem" for this field. Secondly, to unify the diversity of technical language around this field, to differentiate the different disciplines that contribute to,say, "statistical thinking" if you wish - this includes the myth of math and clarifies what place of maths is in stats".

In this field there are several approaches, sources of problems, technical siglas and jerga, contributing disciplines plus a mathematical mixture of sets theory, logics, calculus , and models to represent datasets -obtained from research- and procedures to test and compare results. There are deep disagreements between experts.

For example there is a confrontation between "frequentists" and "bayesian" approaches, another for "linear" and "non-linear" models, another for "parametric" and "non-parametric" models. But all of them use frequences from the start when measuring chances and also use tests of subjective tests about "likelihood", "confidence" and the like. They tend to hide the open discussion of their premises and the limits of their particular models and this effects the quality of their knowledge and applications.

For this reasons I believe your book may help doctoral students to grasp the subject if presented in a simple way, and to understand that their measured datasets may be handled with simpler methods. Many contributions to statistics have been produced by researchers from non-mathematical fields, as many science advances were developed before statistics appeared.

I would be glad to contribute a grain to your task with a short essay that seeks to identify the main practical-theoretical fundamentals of statistical approaches, using simple math simulations. Thanks, emilio

Michael -

Interesting question and answers.  Two terms that come to my mind are "big picture," and "problem solving."  To 'think statistically' is to think logically. The building of an understanding of mathematical concepts is essential. But many statisticians go into a lot of detailed mathematics while going 'off on a tangent' because they are not really addressing the essential issues. I have often seen a gap between theory and practice because the subject matter expert and the statistician did not communicate well. That can happen to any of us, especially in a forum such as ResearchGate where technical questions do not contain all the necessary information, and people trying to answer have different backgrounds, but that is OK, and even desirable, because it is just a start. But we all need to keep open minds.

"To think" as an expert in statistics, I would hope, would really mean to think as an expert in statistics should think, which means understanding the problem to be solved, and the context (big picture).  As in other fields, many may make a distinction between theoretical work and application, but the two need to be linked to be meaningful. When an application is done 'cookbook-style,' this may be worse than useless. And many statisticians are guilty of perpetuating less-than-helpful implementation. Many, for example, fuss over the best way to estimate a p-value, when an isolated p-value is misleading.  So one thing I think we all need to concentrate on is the interpretability of all that is done. 

Students of statistics, and practitioners,  whether statisticians or other subject matter experts, can learn to think conceptually, and learn how to interpret the problem, plan, and interpret results, all in an overall context.  Even in the physical sciences and engineering, and for many statisticians, such experts may need some help with mathematical details, but there needs to be an overall understanding of concepts.  That is essential.

I particularly agree with much of what Jochen said. 

The key to good results is often collaboration, so I see your idea as one to make it easier for various subject matter experts and statisticians to speak a common language, avoiding miscommunication. That is not easy, as it is often difficult to 'put yourself in another person's place' (and can only be done to a certain extent, no matter how hard we try).  Rude/disrespectful wording, fragile egos, and aggressive behavior are problematic.  This can seep over into our writing, so an approach that fosters the respectful  communication of ideas is essential.

Often, this is not easy. 

In my experience, the importance of knowing the basics as to how things work cannot be overrated. I saw subject matter experts and statisticians alike, in practice at an agency, whose purpose, though some apparently did not recognize it, was to produce Official Statistics on the energy industry, who did not try to understand the basics. They might start with a misconception and apply some method, cookbook-style,  and then as things went wrong, they would exacerbate the problem by making dubious 'patches' to their approach,  rather than try to understand the real problem.  Someone would then congratulate themselves, saying how complex their system was, when it was really a mess that could have been better solved much more simply.  This seemed commonplace.

Basically then, I see your "how to think" approach has great merit.  But as an aside, I must say that this, and everything else works better in a spirit of cooperation rather than competition.

And best of all, problem solving using statistics can be fun!  There is a rewarding feeling to be had.

Cheers - Jim 

Whatever I say here, is from the Indian context, based on my experiences alone.

Any textbook of statistics for college students ( In India, mostly textbooks are written as an amalgamation, on the lines of 'Research Methodology+Statistics for xxx') should include introduction to the Philosophy of science. It should be written with the reader's general comprehension level in mind, but is a must. Getting into the world of research without understanding the basics of epistemology should be avoided

The reason why I say this is that for the students that I see around me, Statistics, or for  many of them even Research Methodology, means SPSS only. This situation does not augur well for the future of research.

Dear Michael Filsecker,

I'm not interested on cooperation, because I'm not able to write a similar book, but I'm very interested on read it. Very interesting project.

Mariano