Lecture 1 Flashcards
Acronym
Only Quote Real Sexy Beasts, Cuz Really, Poor Farts Sex Too Slowly, Stopping Errotic Comings & Stunting Mating Good
Rule 1
OBJECTIVES
-State your objectives before you start your project
Rule 2
- Many questions can’t be answered
- Biological hierarchy Increased knowledge as go down.. We know our own DNA etc but not much about biosphere… As we move up into the larger biosphere (ex salt marsh) - mgmt and restoration is difficult because no base data is known
Rule 3
- Just because a variable is measureable doesn’t make it relevant to your study
- May have lots of equipment, but if choose to make measurements, then make sure they answer what is outlined in the objectives
Rule 4
-Make and effort to collect sufficient data for a proper statistical analysis - number of replicated (n) ~10-20
For monitoring sites -> Usually means a trade off between the number of sample sites and the number of replicates
For experiments -> Usually means a trade off between the number of treatments and the number of replicates
ex… 10 sites X 10 reps/site = 100 SU…….. 20 sites X 5 reps/site = 100 SU
Rule 5
- Balance sampling and experimental designs if there is no reason not to
- providing same number of replicates or collecting same number of samples - makes it simpler for analysis later. BUT if stratifying - areas should recieve more attention and more samples (larger areas) - This shouldn’t be ignored. Need to be proportionally allocated.
*Not always appropriate to have a blaanced design.
Rule 6
- Always use controls in experimental studies
- If measuring a response ex regrowth after fire - then it’s not always clear what the control is.
- ex… Nitrogen in soil of plotted plant - wouldn’t make it zero because then the plants would die. Would put in appropriate amount and then raise it.
Rule 7
-Get a representative sample of the study population which is unbiased (=random)…
*This allows generalization to the population of interest. Going from particular to local population to all other populations that are similar but you have not studied them.
*Inductive reasoning from study population to population of interest. - Inferential statistics - the creation of confidence interval because of design.
Rule 8
- Conduct a preliminary survey
- Reason is to become familiar with sampling and sample units. - Can turn into pilot study if preliminary work hadn’t been done in order to understand characteristics of what is being studied.
Rule 9
- Adjust the sample unit (SU) size to fit the organism sampled
- This recognized that organism can very in size and would not be sampled in same fashion… Look at slide..
- With respect to slide.. don’t sample throughout entire plot (sample according to size)
- Sample unit is outline of square… Then within it have the different size sample areas to represet the species being samples and their corresponding sizes… ex. Moss sample area is smaller than the tree sample area.
Rule 10
- Use stratified sampling
i. e., divide the sample area into smaller areas (strata) based on geography and density of organisms - based on percieved homogeneity. - Ex High tide to low tide.. stratify to create areas of homogeneity for sampling purposes. (stratification)
-or a blocked experimental design..
-ie., divide the experimental area into smaller areas (blocks) based on percieved physical gradients. (yellow line) eg location on slope or temperature in greenhouse. - Recognize areas of homogeneity and establish them as blocks (that share the homogeneity)
Rule 11
TEST ASSUMPTIONS OF PARAMETRIC STATISTICAL ANALYSIS
(in order of decreasing importance)
(first three are part of the statistical analysis and must be met before inferential stats can take place)
- Homogeneity of Variance
- Independence
- Normality
- Randomness - deals with bias… Is supposed to not exist and so would need to be addressed
- randomness is part of the study design and the statistical analysis assumes there is no bias
- randomness is assessed as part of the design (e.g., simple random sampling)
If ccan’t meet requirements
- Transform the data… Log10 is common - if works, can proceed using transform data
- Unse non-parametric statistical analysis (ie.. distribution free) - Requirement is that if data is squewed, all data must be changed to be squewed in same way
- resmapling technique - attempt to normalize distribution from squewed dataset
Rule 12
- Should avoid using a statistical shotgun approach to get the answer you want - because have powerful stat software, try to resist rejecting null hypothesis until get desirable result.
- Should create a dummyset to reflect what you’d expect from experiment and then further put effort into determining type of analysis appropriate from data you have.
Rule 13
- Before you start, decide on the number of significant figures to record
- Collecting finer data can cause you to reject null hypothesis just from changing scale of measurements
*only applies if you are part of or continuing an existing study.
Rule 14
-An ecological estimate is of dubious value without some measure of contributing errors - (errors=variability in dataset)
Rule 17
- Be aware and check the mathematical model being applied.. This is especially true with statistical software programs
- Desirable to check what equations are and report them.. Need to make available, because would be useful for people (and you) to test.
