First 1231 words of the document:
AQA A-L evel Biology How Science Works Glossary
Types of Variable Types of Data
Confounding Variables Data
Dependent Variable Any variables that could also affect the (measurements,
The variable you measure, to see dependent variable. Confounding singular datum)
how it is affected by the variables should be
independent variable. controlled in a fair test. Quantitative or Qualitative or
Numeric Data Categoric Data
Independent (numbers) (words)
Variable Confounding variables that Continuous Data Discrete Data Ordered Data Nominal Data
The variable you choose independent variable are kept constant can have any value only whole numbers can be ranked can't be ranked
to change, to see how it (controlled) during the e.g. 7.34, -294.6, 2x10
e.g. no. of atoms e.g. small, medium, large e.g. male, female
affects the dependent variable. experiment. If you can't control a variable
You may also measure it when (such as weather in a field investigation),
you change it. you should at least monitor it Errors
Random Errors Bias
Experimental Design Inaccuracies due to mistakes, poor When the observer chooses some
technique, or random variation. results and ignores others,to support a
Controlled Experiment (FairTest) Random errors are very common, but particular view. Also called cherry
When all relevant variables are controlled, so that observed changes in the can be improved by taking many picking.
dependent variable must be due to changes in the independent variable. replicates. Data with a small random
Anomaly or Outlier
Control Experiment (Control) error is said to be precise.
A measurement that falls far outside
An additional experiment designed to eliminate alternative explanations for the the expected range and is therefore
main experiment and so show that observed changes in the dependent variable Systematic Errors probably due to experimental error.
must be due only to changes in the independent variable. Inaccurate measurements in one Anomalies should be rejected, since
d i re c t i o n o n ly, d u e t o p o o r they skew the mean, but it is very
Control Group calibration or poor technique. difficult to distinguish between
A group or sample treated in the same way as the experimental group, except Systematic errors can not be anomalies and normal biological
for the factor being investigated e.g. a placebo group in a drugs trial. By improved by taking more replicates. variation.
comparing the results for two groups it can be shown that observed changes in Data with a small systematic error is
the dependent variable must be due only to changes in the independent variable. said to be reliable.
Placebo Zero Error Ensuring that a measuring instrument
A dummy pill, injection or treatment that has no physiological effect (e.g. a sugar A particular kind of systematic error, gives correct readings by fixing known
pill or saline injection). Used in a clinical trial to allow for the placebo effect - where the instrument does not return to points then construction a scale
the observation that symptoms can improve when patients believe they are zero. between them.
RCT Simple Analysis
The best experimental design for a drug trial. RCT stands for Randomised Standard Deviation (SD)
Controlled Trial, or in more detail, a Randomised, Placebo-Controlled, Double- Replicates A measure of the dispersal of the replicates
blindTrial. This design ensures that the trial is valid free from bias. Repeats of a measurement. about the mean. In a normal distribution 68%
! Randomised means the study and control groups are allocated randomly
of the replicates will be within 1 standard
! Placebo-controlled means the study group (taking the drug to be tested) is Raw Data deviation of the mean, and 95% will be within 2
compared to a placebo group (who are given a placebo). The original measurements or standard deviations of the mean.
! Double-blind means that neither the subjects nor the investigators know re c o r d i n g s b e f o re a ny
who is in the study or placebo groups.This avoids bias. manipulation or processing.
Standard Error of the mean (SEM)*
Protocol A measure of the uncertainty, or error, of a
A method or technique that has been shown to produce valid and reliable Mean or Average calculated mean.The smaller the standard error,
results. T h e m i d - p o i n t o f t h e the more reliable the mean.
Hypothesis = sum of replicates / N
A suggested explanation of observations or results that can be tested. Also 95% Confidence Interval (CI)*
known as a scientific hypothesis. A good hypothesis can be used to make Another measure of the error of the mean.We
The highest and lowest can be 95% confident that the true mean lies in
replicates, or the interval the range (mean ± CI). The top and bottom of
Quality of Data between them. this range are called the confidence limits.
TrueValue Measurement that are close to the
The real value of a measurement, if it true value. Statistical Analysis
could be measured with no errors at
all. Valid Data Correlation (orAssociation) Null Hypothesis*
The best quality data, i.e. data that is When one variable changes with The statement that is tested by a
Precise Data precise and reliable and obtained another variable, so there is a relation statistical test. The null hypothesis is
1. Measurements that give similar from an unbiased , controlled between them. However it may not be fixed for each test, but always says that
values when repeated. The experiment that addresses the stated a causal relation. The strength of a there is no difference or no
replicates therefore have a small aim. Valid data is assumed to be correlation can be measured using the association. The null hypothesis has
range. Pearson or Spearman test. nothing to do with a scientific
2. Data measured on sensitive hypothesis.
equipment with a suitably fine scale, Evidence Causal Relation
e.g. 20 mm is more precise than 2 Any data or observations that are used When changes in one variable cause
cm. the changes in another variable. Can P-value*
to support a particular hypothesis. only be shown by a controlled The result of a stats test,expressed as a
Reliable Data Anecdote experiment. probability. It is effectively the
Findings that can be repeated. This An observation or story from real life. probability that the null hypothesis is
includes by the original investigator; by Anecdotes are not evidence and StatisticalTest*
true. If P<0.05 then we reject the null
other scientists; by other techniques; cannot be used to support a Something that tests whether
hypothesis,otherwise we accept it.
or those that agree with secondary hypothesis, but they can be useful to observed differences or associations
sources. suggest a new testable hypothesis. are significant,or just due to chance.
HGS Biology * = A2 only NCM 01/09