Test 3 Flashcards
Goals of ethics review
any human research must go through the IRB
- Prevent gross ethical violations
- Correct for blind spots
History of US regulations
1974 National research act protection of biomedical and behavioral human subjects
1979 Belmont report
1991 Common Rule–set of federal regulations
Belmont Principles
- Respect–informed consent, time, procedures, risks and benefits, voluntary, NO COERCION
- Beneficence–research must do good for someone. Medigate risk and justify it. Confidentiality
- Justice–risks and rewards. Not equal. researchers reaped benefit from vulnerable people’s risk
IRBs
every IRB must be federally registered
- 2 kinds. A promise to govt and set up our own IRB
- if only the first, you can use another institution’s IRB
IRB membership
- represent all three divisions. natural, social, humanities
- community members in our case the bishop and the Sanford employee.
IRB duties
- review every proposal. not everything is research.
- only wide generalized knowledge, not classwork, history, stuff for augie.
IRB guidance
regulations–who is the primary investigator?
policy
judgement
IRB process
-intial proposal (form)
submit all materials like flyers and informed consent forms
-exempt, expedited and full
expedited is minimal risk encountered in daily life
-reviewers
checklists and need some mods
further requests to primary investigators
-final decision–can be denied
-need to follow up. Have to say what happened and report any adverse effects. must submit another request to continue the study
Guidelines for figures
- simplify without falsifying data
- graph/table
- brevity and clarity
- note prior conventions
Tables
-need specific data points
-useful for exact comparisons
-stand on its own
-show data and possible manipulations
%s, total, means, averages
-define abbreviations at bottom
-units at top
-limit number of things in the table
-align stuff
-dont go crazy with decimals
-verify all data
-accurate use of symbols
Bar Charts
- less numerically specific
- examine differences rather than trends
- compare size, magnitude
- bars need to be wider than the spaces between them
Line graphs
-not numerically specific
-demonstrate movement, changing trends
-generally over time/concentrations
-regular intervals with two axis and units
don’t do line patterns, vary symbols.
Graph parts
- legend or key
- label at the bottom
- units on axis
- use color and symbols to your advantage!
Figures general guidelines
- Is a figure needed?
- no top title
- limit 3-5 curves, 6-8 bars
- label axis with units
- start scales at 0 otherwise be clear about what youre doing
- legends and keys
- should stand alone
Referring
- every table or figure should be referred to in the text
- whats interesting about the figure? not just see results in blah blah
How do we determine experimental unit and the replicate?
- experimental unit stands alone
- replicates are related in some way
experimental unit
- physical entity which can be assigned a treatment at random
- can receive either the treatment or not
Pseudoreplication
-treating multiple measurements on the same units as if they are measurements on independent units
What is replication?
the smallest experimental unit to which a treatment is independently applied
treatments must be repeated or independent experimental units in order to avoid pseudoreplication
-how many depends on the natural variation of the data set
MUST HAVE
- replication
each treatment tested on more than one experimental unit - randomization
experimental units are allocated to treatments at random
Example: grass was assigned to 3 different trays and 54 plants were planted
- sample is 3. n = 3.
- randomize by moving around in the growth chamber
Principle component analysis
- data reduction technique
- only multivariate technique we will look at
- reduce to see what mainly controls data variation
Data reduction
- sum of data with many p variables by smaller set of k derived (synthetic, composite) variables
- balancing act between clarity of representation and ease of understanding. If you oversimplify you may lose relevant data
- Pearson and Hotelling in ecology developed
- most widely known and used “standard” tests
- matrix of n objects by p variables which may be correlated
- m components ranked by variability
Principle components
- ranked by how much variability they explain.
- NOT related to each other, not linearly correlated
- composite variables, combo of variables
How principle components are found?
- objects represented by a cloud of n points in multidimensional space
- centroid is defined by the mean of each variable
- variance is the average standard dev
- linear correlation = covariances –> correlation with the same units and scale
Geometric Rational
-rigidly rotate axes of p dimensional space to new principle axes that:
explain the biggest part of the variance
PC1 highest variance
PC2 second highest variance
-covariance between each pair of PC axes is zero. NOT related
-in practice use more than two variables
-PC1 is direction of max variance and least squares line of best fit
Principle Components Analysis Assumptions
-MUST be linear
-If non linear the P axis won’t work
used in ecology often
Two factor ANOVA with equal and unequal replicates
-simultaneous analysis of effect of more than one factor
-Factorial analysis of variance
ADVANTAGES:
-one set of data is enough (multiple one way is too much error)
-economical
-test for interactions that are often biologically signifigant
Two Factor ANOVA assumptions
- normal distribution
- equal variances
- randomization
- no heteroscedastic
- departures from these, if p is close to alpha, then conclusions are suspect
- try transformations!
3 models
- Model I (all factors fixed)
- Model II (all random factors)
- Model III (some fixed and some random)
For a two factor ANOVA you have six hypotheses
- Ha1 Factor one will have an effect
- H01 Factor one will not have an effect
- And same for two more sets
Mathy ANOVA things
grand means sum of squares total sum of squares cells cs of s - total s of s -calculate which is more important and causes the most variation
How to read a two factor ANOVA correlation graph
- slope = same means no interaction
- height on graph correlates to the effect, symbols for one compare horizontally and vertically
Choosing journal
- choose with care
- choosing highest impact factor isn’t always in your best interest
Journal writing guidelines
- precision economy, clarity
- check grammar and spelling
- see journal for voicing
- be careful with paraphrasing. Need to have good citations
General writing
- start with intro and methods.
- write design as you go. record data
- refine methods
- analyze and interpret results
- discussion
- title and abstract
- self revise 2-3 X before you ask others to read
- ask for peers you know would be good to read your paper
Submission
- cover letter
- register with journal
- texts and figures and legends all separate
- editorial review takes 2 weeks to 2 months
- specify the reviewers you do or do not want in you cover letter
- peer review 2-6 more months
Submission outcomes
-accepted outright
-minor revision
-major revision
additional revision / experiments
-rejected :(
rm
reset R
ls
list
setwd
set working directory
getwd
get working directory
read.csv
read in a csv file
names
retrieve in data names
head
see begin of data frame
dim
see dimensions of object
str
structure of object
summary
exploring data
[]
used to call specific points [row,column]
subset
pull out specific value
$
use with data frames
==
comparison
&
and
|
or
!
not
!=
not equal
aggregate
split objects and compute summary stats for each
tapply
apply function over ragged array
mean
finds mean
sd
standard deviation
median
find the median
