Experimental Design for the Life Sciences《我做的读书笔记》

原文在：

http://lijuan.yo2.cn/2008/07/22/reading-experimental-design-for-the-life-sciences-chapter1-why-you-need-to-care-about-design/

http://lijuan.yo2.cn/2008/07/23/reading-experimental-design-for-the-life-sciences-chapter2-starting-with-a-well-defined-hypothesis/

http://lijuan.yo2.cn/2008/07/23/reading-experimental-design-for-the-life-sciences-chapter3-between-individual-variation-replication-and-sampling/

http://lijuan.yo2.cn/2008/07/25/reading-experimental-design-for-the-life-sciences-chapter4-different-experimental-designs/

还有一个flowchart在：

http://lijuan.yo2.cn/2008/07/25/reading-experimental-design-for-the-life-sciences-flowchart/

Chapter 1 Why you need to care about design

* Experimental design is more about common sense biological insight and careful planning.

* Poor experimental designs waste time and money also have ethical issues.

* Every statistical test have slightly different assumptions so it is essential to decide in advance how you will ***yse your data when you have collected them.

* The two major goals of designing experiments are to minimise random variation and account for confounding factors.

下面是新手们经常有的两个认识误区：

Myth1 It does not matter how you collect your data there will always be a statistical 'fix' that will allow you to ***yse them.

Myth2 If you collect lots of data something interesting will come out and you'll be able to detect even very subtle effects.

Chapter 2 Starting with a well-defined hypothesis

A hypothesis is a clear statement articulating a plausible candidate explanation for observations.It should be constructed is such a way as to allow gathering of data that can be used to either refute or support this candidate explanation.

For example:

1. Questions: why does chimp activity vary during the day?

2. Hypotheses: Chimp activity pattern is affected by feeding regime.

3. Predictions: The fraction of time that a chimp spends moving around will be higher in the hour aroung feeding time than at other times of day.

Make sure that your experiment allows you to give the clearest and strongest evidence for or against the hypothesis.

Make sure that you can interpret all possible outcomes of your experiment.

Pilot study: Exploration of the study system conducted before the main body of data collection in order to refine research aims data collection and ***ysis techniques.

Correlational study's advantages:

* we handle them with much less time;

* We do not affect other functions;

* We are dealing with biologically relevant variation

Manipulative experiment's advantages:

* without third variables;

* without reverse causation;

A more efficient approach might be to begin with a large correlational study to see which factors seem to be important. Once potentially influential factors had been found manipulative studies could be used to confirm and refine these findings.

There is no perfect study but a little care can produce a good one instead of a bad one.

Chapter 3 Between-individual variationreplication and sampling

Whenever we carry out an experiment we are trying to find ways to remove or reduce the effects of random variation so that the effects that we care about can be seen more clearly.

Replication involves making the same manipulations and taking the same measurements on a number of different experimental subjects. Replication is a way of dealing with the between-individual variation due to the random variation that will be present in any life science experiment.

Replicate measures must be independent of each other: Techinicallly statisticians talk about there being no correlation between the deviations of individuals within a group.If this is the case this will have the important effect of meaning that if we examine a group of independent individuals their deviations will tend to cancel out and the mean of the sample will close to the mean of the population.

Pseudoreplication is a problem that has to be addressed by biologists and not by statisticians.

Accept that sometimes pseudoreplication is unavoidable so the key if you can't replicate fully is to be aware of the limitations of what you can conclude from your data.

Number of replicates: It should be big enough to give you confidence that you will be able to detect any biologically meaningful effects that exist but not so big that some sampling was unnecessary.

* Educated guesswork: reference on previous similar studies.

* Formal power ***ysis: statistical power is the probability that a particular experiment will detect a difference assuming that there really is a difference to be detected. (there are many computer programs that can help you)

Randomisation simply means drawing random sanmples for study from the wider population of all the possible individuals that could be in your sample. Proper randomisation means that any individual experimental subject has the same chance of finding itself in each experimental group.

The power of an expeiment will be affected by three main things: the effect size the amount of random variation and the number of replicates.

Chapter 4 Different experimental designs

The control group must be that it differs from the treatment group in no way except for the treatment being tested.

A blind procedure is one in which the person measuring experimental subjects has no knowledge of which experimental manipulation they have experienced or which treatment group they belong to.

In experiments with humans we may use a double-blinded procedure in which the experimental subjects too are kept ignorant of which treatment group they belong to.

The procedures are design to remove the perception that unconscious bias might taint the results of a study.

后面讲的paired-design cross-over design split-plot design就不如直接看统计书了。

Chapter 5 Taking measures

Calibrate your measuring instruments(including human observers.)

Adopt clear definitions to reduce subjective decision-making during measurement taking.

Watch out for observer drift intra-observaer variability and inter-observer variability.

Watch out for observer effects where measuring a system influence its behaviour.

Recording data effectively is a skill that you must acquire.

1. Don't try and record too much information at once;

2. Beware of shorthand codes.

3. Keep more than one copy of your data.

4. Write out your experimental protocol formally and in detail and keep a detailed field journal or lab work.

5. Don't overwork.

Chapter 6 Final thoughts