To start with, this is what I use in my course:

Research Methods: Bordens & Abbott's Research Design
and Methods (soon to be in 8th edition).  It is it broad in its
coverage of methods, design, as well as having a three
chapters on statistics (one on descriptive, one on inferential,
and one on multivariate techniques).  It's coverage is somewhat
uneven but I try to compensate for deficiencies in my notes
and lecture.  When this course is taught first in the methods-stat
sequence, I refer to informally refer to correlations, means, and 
the null hypothesis tests for them but (a) always in the context of 
a specific example and (b) without computations to highlight what 
aspects of the data we need to focus on (e.g., in an RCT with a 
drug and placebo group for treatment of depression, what should 
our initial expectation be about the mean BDIs for these groups and
what is it we want to see about the mean BDIs in order to reject 
that initial position and to claim that we have evidence for an effect 
of medication on depression).

Statistics: Gravetter & Wallnau (6-7th ed) supplemented
with Dretzke's Statistics with Microsoft Excel and examples
of SPSS analyses but using syntax instead of the menu system
which comes in handy for repeated measures ANOVA.

NOTE:  One should always teach that the use of Excel in statistics
is only  appropriate for limited analyses of the data and even these
should be double-checked by a "real" statistics software.
There is an extensive literature on the problems with Excel
in statistical analysis with Bruce McCullough having written
several papers on the topic.  Here is a link to a two page
article by McCullough that can serve as an entry point to the 
literature:
http://www.forecastingprinciples.com/files/McCullough.pdf

On Thu, 15 Apr 2010 06:09:01 -0700, John Kulig wrote:
>Marie et al
>
>Not that I want to pitch MINITAB commercially, but I am always 
>amazed at preconceptions. MINITAB has cronbach's alpha & factor 
>analysis under Multivariate stats, multiple R, binary ordinal & logistic 
>regressions under Regression, all anova options, etc. I would find it 
>amazing if the go-to stat package for statisticians would not have 
>these common procedures. 

John's points are justified but I would only point out that in the business
world, Minitab is probably unheard of.  SAS and to an increasing degree
"IBM SPSS" (it's new name after taking on PASW for a short period
of time) have been used in the business community as well as in medical
research.  One can imagine that IBM will aggressively promote the use
of SPSS in their markets and students with some knowledge may have
a edge "jobwise" relative to students without such knowledge.  A business
might be more interested in a person who knows how to do statistics in
Excel than in minitab.

I need to go off on something of a tangent here on the issue of whether
one should use Cronbach's alpha (implying that if we teach about it,
we need to be more cautious about what we say about it).  Over on
the Structural Equation Modeling (SEM) mailing list, it is a common
activity to flog those who use Cronbach's alpha.  Here is a recent post
to the group that provide references to current thinking about the
nature of Cronbach's alpha:

|From: Structural Equation Modeling Discussion Group

|[mailto:[email protected]] On Behalf Of Cameron McIntosh

|Sent: Friday, April 09, 2010 10:37 AM

|To: [email protected]

|Subject: Re: Negative Cronbach Alpha

|


| Hi Fatma,


| At some point, your student should have a look at the following. Not

| that I have a vendetta against alpha or anything, just trying to assist

| with the gradual evolution of the field. :)

| 

| Sijtsma, K. (2009). On the use, the misuse, and the very limited

| usefulness of Cronbach's alpha. Psychometrika, 74(1), 107-120.

| 

| Bentler, P.M. (2009). Alpha, dimension-free, and model-based internal

| consistency reliability. Psychometrika, 74(1), 137-143.

| 

| Green, S.B., & Yang, Y. (2009). Commentary on coefficient alpha: a

| cautionary tale. Psychometrika, 74(1), 121-135.


|

| Yang, Y., & Green, S.B. (2010). A note on structural equation modeling

| estimates of reliability. Structural Equation Modeling, 17(1), 66-81.


|

| Ogasawara, H. (2009). On the estimators of model-based and maximal

| reliability. Journal of Multivariate Analysis, 100(6), 1232-1244.


|

| Huysamen, G.K. (2006). Coefficient alpha: Unnecessarily ambiguous,

| unduly ubiquitous. SA Journal of Industrial Psychology, 32(4), 34-40.

 

If one uses Cronbach's alpha regularly, they really should be familiar
with this literature and what the alternatives are.

>But I understand that the majority of psychologists choose SPSS first. 
>A few years back our IT people pushed for more MINITAB and the 
>psych depart and a few others fought to keep SPSS on the network, 
>though the arguments did not revolve around specific procedures that 
>were used, just personal preference. I always joked that MINITAB's 
>only fault was the name. It might be better called MAXITAB or 
>POWERTAB or something like that.

I think that a more relevant inadequacy of Minitab is that it is not in
common use in either the business or medical communities, both being
lucrative markets for software companies.  "IBM SPSS" is probably
going to be used less and less in undergrduate education as the cost of a
license will continue to increase (even academic licenses have to keep
pace with the the cost of business licenses).  If I am not mistaken,
at a number of colleges a license for SPSS is now more than that for
SAS which is a reversal of their historical relationship.

What is the solution?   As colleges and universities try to hold down
costs, such as the costs involved in software licenses, I suspect that
there will be major shift to using a program like "R" because it is free
and its development will should make it accessible to people without
much stat background (today there is R commander but I think that
needs more work).  I think it will be the software that serious statistician
will use because new statistical procedures will be implemented in it
before it is implemented elsewhere (which is why a number of programs
like SPSS and others allow one to call R from within the program and
use its analytic procedures).

So, I would suggest folks should probably start getting familiar with R.
There are a number of resources on the web as well as commercially
available sources (e.g., books on R for different level of knowledge).
Ultimately, I think it will be an economic decision that decides which
statistics software will be used and taught.  Don't be surprised if in
the near future schools move from Microsoft Office to Google Docs
and related free software.

-Mike Palij
New York University
[email protected]






---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here: 
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=2004
or send a blank email to 
leave-2004-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to