Paging the Science / Biology / Ecology types

Can any of you suggest some decent research methods / stats type books for
this sort of thing?

 

paging the pixies. Retail opportunity...

 

Undergrad dissertation. Looking for quick guidance on stats in excel and
which stats do what in book form. I'm hoping to find some patterns in data
for bats using various kinds of man made bat box.

 

My advise would be to leave well alone. Just stick to reporting stuff
with maths you can understand, and state quite clearly that you'll
leave more detailed stats analysis to the professionals.

If you do try to do it yourself, you're building a rod for your own
back, as any real statistician will be incapable of looking at your
report without tearing it to pieces. It's not personal, they just
can't help it. Maybe it's genetic, but put three statisticians in a
room with a poorly-designed study report and you'll get at least five
distinct "correct" interpretations.

--
_______
..'_/_|_\_'. Ace (b.rogers at ifrance.com)
\`\ | /`/
`\\ | //' BOTAFOT#3, SbS#2, UKRMMA#13, DFV#8, SKA#2, IBB#10
`\|/`
`

 

One of the friendlier ones is this ..

Dytham, C. (2003) Choosing and Using Statistics - A Biologist's Guide.
2nd ed. Oxford: Blackwell Publishing.

It's a very accessible stats book which has a handy key at the
beginning which will guide you to which stats to use for what type of
data you're collecting. Chances are that your library will have a
copy.

The OU also do a reasonable guide on research methods and experimental
techniques ... I can't think of the exact title but again, any half
decent uni library will have it in the research methods/stats section
(somewhere around the 570s).

As for using Excel, hhmm, whilst it's ok for some simpler stats stuff,
you may find having access to something like Minitab or SPSS (1)
useful. Does your uni do a discount software buying thing at all? if
not, drop me an email, I may be able to help.

What year are you in? Hopefully not final if you're starting to look
at stats for a dissertation now

HTH

1 - yes I know you proper stats people hate them

--
Elly - A Pixie doing a PhD
ZX9R-E1 - <Giggles>
Spike - FZ400 - It's dead Jim!
MRO#32 ibW#25 BoTAFOT#46 BoTAFOF #46 GP#1 UKRMRM#00 TWA#3
DFV#15
http://www.garagepixies.co.uk
elly at garagepixies dot co dot uk

 

This is free and written by the master of data analysis:

http://www.edwardtufte.com/tufte/dapp/

 

R is free: http://www.r-project.org/

 

I'm sure you've probably got plenty already, but I may be able to
point you in the direction of some interesting papers on bat stuff and
use of roosting sites.

Unfortunately (for students) most dissertations for applied science
degrees will require a good level of detailed stats analysis to be
carried out. Mine was a complete nightmare and involved using
principal component analysis (amongst other things) to assess
underlying environmental gradients in my data ... it was 'fun' to say
the least.

This is the part I'm dreading with my research ... as my project
crosses several levels of above/belowground ecology, I'm getting all
sorts of conflicting advice, some of it from my supervisors ...aargh


--
Elly - A Pixie up to her elbows in soil
ZX9R-E1 - <Giggles>
Spike - FZ400 - It's dead Jim!
MRO#32 ibW#25 BoTAFOT#46 BoTAFOF #46 GP#1 UKRMRM#00 TWA#3
DFV#15
http://www.garagepixies.co.uk
elly at garagepixies dot co dot uk

 

We've had lectures on SPSS but according to my supervisor it's not necessary
to get into that level of statistical pain. More looking at the fairly
simple stuff you can do in Excel - t-tests and the like. I already have a
subsidised copy of SPSS but I've no desire to start wrangling with getting
it activated on a new PC or with its arcane interface.

I'm not one for starting things early...

 

Don't use Excel:

Computing Algorithms for Basic Statistics
Excel Uses Poor Algorithms To Find The
Standard Deviation (See Help screen for
STDEV shown above)

Excel Defines The First Quartile To Be The
Ordered Observation At Position (n+3)/4

Excel Does Not Treat Tied Observations
Correctly When Ranking

Regression Computations Are Often Erroneous
Due To Poor Algorithms (See below)

In addition Excel, usually displays many more digits
than appropriate. (See histogram and paired t-test output
shown above.)

Finally, Excel has major and documented
difficulties with its regression procedures.

Regression in Excel

Does Not Treat Zero-Intercept Models
Correctly

Sometimes Gets Negative Sums Of Squares

Does Not Handle Multicollinearity Correctly

Computes Standardized Residuals Incorrectly!

Displays Normal Probability Plots That Are
Completely Wrong!

Makes Variable Selection Very Difficult

www.cs.uiowa.edu/~jcryer/JSMTalk2001.pdf

 

I'm pretty much there already. It's only an undergrad dissertation...

I'm almost grateful for the stats module I paid no attention to in A-level
maths. Still, barely any of the stuff we covered is used at degree level. I
feel sorry for the luddites and those who didn't do maths past GCSE who are
grappling with this stuff...

 

I'm sure you're right, *but* if that's what's been recommended to me by my
supervisor, who will presumably also be marking it, I can't see any sense in
using SPSS when I don't have to. Given that more time I might go down that
route, but it's not especially practical now.

 

LOl, so was mine, but the refs still ran to 6 pages of primary
literature. You (and your supervisor) sound far more relaxed about it
all than I was.

That would have been me then ... got a CSE grade 4 in maths *way* back
in 1983(ish), partly thanks to inept teaching (1) at the age of 9/10
that gave me a huge confidence problem with anything maths related.
Thankfully, the access course I did prior to starting the degree got
me back on track with maths, and the stats training that we received
at the University of Northampton was absolutely excellent.

Just goes to show that you can teach an old dog (2) new tricks!

(1) Thank you Mrs Sewell, it took a very long time to undo what you
managed to do in less than a year.
(2) Quiet at the back there.

--
Elly - A Pixie doing a PhD
ZX9R-E1 - <Giggles>
Spike - FZ400 - It's dead Jim!
MRO#32 ibW#25 BoTAFOT#46 BoTAFOF #46 GP#1 UKRMRM#00 TWA#3
DFV#15
http://www.garagepixies.co.uk
elly at garagepixies dot co dot uk

 

Blimey, that's going some. I reckon I'd struggle to find that amount of
stuff to be honest. That said, I've only lately properly grasped how to get
at journal articles online, after 2 years of never being able to find a
thing. If you think you have anything useful I'd be very grateful if you can
send it across BTW. My reply address is valid.

I'm probably relaxed because I'm used to working close to the wire, and I
think my supervisor has given up hoping to change things... Always been like
it with academic stuff.

It's amazing just how much difference a bad teacher can make.

 

That was about normal for our cohort. One thing I'm very grateful to
UoN for was the emphasis placed on using journal papers from year one.
I take it you do have access to the wonder that is 'Athens'? if you
do, 'Web of Science' is an excellent source for finding papers and
also has a handy citation thing, so you can easily find related
papers. Scopus is also great for that if you have access.

I'll see what I can find.

Heh, most of my best stuff was produced in a mad flurry of activity
merely days before hand-in dates. A hard habit to break, for sure.
My dissertation was written over a period of 2 weeks, with the bulk of
it being in the final 72 hours (with no sleep) before the due date ...
still managed an A-. I swore that I'd never put myself in that
position again after that ... guess what, I still do it

--
Elly - A Pixie doing a PhD
ZX9R-E1 - <Giggles>
Spike - FZ400 - It's dead Jim!
MRO#32 ibW#25 BoTAFOT#46 BoTAFOF #46 GP#1 UKRMRM#00 TWA#3
DFV#15
http://www.garagepixies.co.uk
elly at garagepixies dot co dot uk

 

*ding*. Data has been given to me and whilst it's probably not ideal for
analysis it's in an area that's doable for a dissertation. A lot of the
other stuff I looked at resulted in naff all appearing during the lit search
phase of things.

Sounds like I'll be looking at non-parametric then. Ta for the help.