BMS650 Flashcards
Why use Controls?
- Reduces & detects confounding & nuisance variation
- Provides a clear answer for comparison so that we can see what a change would look like
- Provides context
TYPES of controls
- Positive (tells you that your experimental system is working the way it should e.g., injecting cancer cells to ensure that the mice get cancers = system working)
- Negative I (tells you what things look like if you don’t
change INDPT VAR e.g., injecting mice with cells that don’t produce cancer) - Negative II (Unperturbed case where the variable
is a component of thing you add) - Method (Doing the method a different way to produce data that supports your point - this produces complimentary data)
- Randomization (if you’re ordering something, you have to be aware that sometimes by ordering, you have biased the data that you will receive and so you have to switch up the order)
- Representative (looking for something that actually depicts the population of interest you’re studying better that the one you’re currently looking at)
- Blinding (having no idea about the samples being processed so that you don’t pay more attention to the ones that are positive and you decrease your threshold for detecting something to the ones that should be negative)
What do controls tell you?
-if your experiment works as expected
-if differences are due to treatment
-the extremes (values) you can expect (and need to detect) - e.g., how bright of a light you should be looking for especially when there will always be background noise/light or what the lowest light should look like
-whether your method is valid
-whether your reagents work as expected
-whether your results apply to a larger population or a subset (sample)
FIXED vs partial control
- Fixed: the variable is isolated, static and unchanging (e.g., biological sex, maybe Genotype – fixable compare mutant to isotype) or, random (“noise”)
-partial control = DIET (e.g., giving the same mouse diet but it’s up to the mice how much they control), TEMPERATURE, LIGHT
An Ideal Experiment
-one cell that behaves identically to other cells and to every other cells in people
- all the samples of the culture are identical and everything behaves the same
-differences between samples are indistinguishable - basically clones since the growth is controlled, you know when it will divide etc
-error bars will be super small
-any samples or treatments that you do, any change that you detect is due to your treatment chosen
Replicates
-Repetition implies stability - the usual situation or a particular situation in a specific setting that is reproducible
-you do the same. thing multiple times to show that you get a similar effect
- if you can show a similar effect multiple times, then you can be relatively confident that the treatment you did has caused the effect
- if you repeat the experiment in a different setting and you see something similar, then the results can be generalized
How many experiments do you need?
power (whether you can find a significant difference) of your test depends on three things …
-magnitude of effect
-test efficiency
-significance criteria p < 0.05
How does Replication Affect Analysis
- more samples increase the precision of the sample mean
-the number of times you repeat an experiment affects the error bars that you get and you report the average
- As your sample size increases, the size of your error bar in confidence interval drops and then stabilizes into a 95% confidence interval
-standard error of the mean (SEM) = taking the means from different experiments and averaging it ti get the mean value over different things - for error bars, you take the standard deviation and divide by the square root of the number of replicas to get much smaller error bars
Descriptive vs Inferential Data
-descriptive - standard deviation and range (lowest and highest points) since they both only require the raw and you don’t have to worry about replicates
- inferential = standard error (it says something about the mean of what you’ve sampled instead of actually describing the population) and confidence interval
Biological Replicates
-taking individual subjects and treating them the same way (same conditions) but each one represents a different biological event or unit of biology
- example = Driving simulation/ rx time
N = 100 people
40% avoid
30% collision
30% clip cow
-therefore, the biological unit here would be the individual tested (as reaction time would vary for everyone)
- but =, if you change it to What effect does alcohol have on reaction time? and you give 50 people water and 50 people alcohol, your N changes
- therefore, number of replicates is determined by the mount if individuals that are getting the same treatment that you’re looking at
- example = What effect does alcohol have on an individual’s reaction time?
question does not need biological replicates - this is closer to technical replicate since you’re only testing one person’s response and so you can’t apply it to others even though you do see changes when you’ve tested that person for example, 10 times
= “scientific anecdote”
Biological Fluctuation - Included in Biological Replicates
-let’s say you have 3 ants and you want to count the amount of legs on each ant - if you and someone else both count each of them and compare then that is 2 technical replicates
-however, the biological replicate would be 3 since we counted 3 ants
- People usually do n=3 so that you can get an average and see how far you’re deviating from the mean
= intrinsic biological variation
=samples the range of differences you might see or measure in a population
=the more variation in a population = greater n
TECHNICAL REPLICATES
-n = amount of times you measure something
-tells you about your ability with measuring and how good you can measure something
- you cn get low variability from your measurements/more precise but it different from your biological population
-intrinsic variation in system, independent of physiology
-High Technical variation = some non-biological
variable responsible
Experimental Replication
perform the entire experiment multiple times
for reproducibility (à MODEL)
Pseudoreplicates
technical replicate fobbed off as biological
-Replicates are not STATISTICALLY independent
of each other but are described as biological
