1	Doing Experiments on Populations What are the ways science generates knowledge (=truth)? With emphasis on biological populations
2	Reading Assignment Reread section 1.4 “Doing Biology” p. 10-14. Figure 34.10 (p.789) presents a type of experiment that may not evaluated statistically because outcomes are + or -.
3	Types of Science Discovery Look for or try something new (not known) Measurement Instrumentation for more precision or accuracy Experiment (hypothesis testing) Comparing groups, interpreting outcomes Theory, logic, mathematics Models of global climate
4	Practice versus science Doctors and engineers use the knowledge generated by science to fix or build things. This is also a creative process as the most knowledge (generalities) have to be applied to the particulars of individuals or circumstances. Science focuses on building our knowledge base and does not emphasize the expected usefulness of knowledge.
5	Generality Science seeks knowledge that applies to all individuals of a ‘type’. All Carbon atoms behave in a certain way, but more refined measurements may be able to distinguish ¹²C from ¹³C, then the two isotopes can be subpopulations of C. In biology populations within a population are often distinguished by gender or age.
6	What is a Variable? Constants don’t ever change, e, π, 5. The word variable is used for attributes that do change, either thru time or among individuals in a population. Some variable take only two values (states), e.g., [YES, NO], others can take any value. A parameter is a value that does not change within the equation used, but changes when the equation is applied to different types, i.e., half lives or doubling times.
7	Types of Variables Measurement variables Individuals are measured on a continuous, ordered scale (all the attributes we talked about in lecture 2 –length, mass, speed, etc. Median, median, variance, standard deviation apply to measurement variables. Categorical variables There are a small number of types and individuals are sorted into types and then counted, e.g., red eyes or white eyes.
8	Populations Individuals within a population are variable with respect to many attributes. Often it is useful to divide a population into subpopulations. For example, humans may be better understood by dividing the whole into groups based on age. We prefer to make statements (hypotheses) about more inclusive populations.
9	Statistics Any value that is calculated from a set of observations is a statistic. A statistic conveys information about the observed values without giving all the observations. Prevalence and Incidence (introduced in lecture 8) are statistics of a disease in a population. Mean and median are statistics that measure ‘central tendency’ -where middle is.
10	Causation & Correlation If we apply an action to individuals and we always get a particular outcome, we say the action caused the outcome. One can also infer causation from observations without applying an action. These cases often involve correlation of pairs of variables observed on a population of individuals.
11	Experiments look for outcomes following actions Do something and observe what happens Can you detect an outcome (what happens)? How long do you wait (for what happens)? Does it work over and over? (Does it happen every time you do it?) Repeatability, Replication, Reproducibility Action by different experimenters Action on different subject at the same time Action on different subjects at different times
12	What motivates one to do an experiment? Innate curiosity and desire to explain why One wants to learn how to do things better Logic applied to a set of ideas believed to be true leads to a predicted outcome (=hypothesis). The materials needed to do the experiment are easily available.
13	Does doing something particular work (better than chance)? The effect (=treatment) being evaluated (tested) needs to be compared: 1) To ‘normal conditions’ – things happen anyway. Thus we need a CONTROL group to which the treatment is compared. 2) To a prediction of what would happen if the treatment tested had no effect –a null hypothesis.
14	The HYPOTHESIS A hypothesis is a statement about what one believes to be true and can be evaluated by observations, i.e., is testable. A hypothesis (statement) can be useful even if it is later replaced by a new hypothesis. e.g., Linnaeus hypothesized that living organisms were divided into plants and animals, but now we recognize 3 Domains as the major divisions of organisms.
15	Multiple Explanations If a a hypothesis leads to a prediction and the prediction is confirmed, it does not mean that any other hypothesis is not true. More than one hypothesis often leads to the same prediction. It is desirable when hypotheses are mutually exclusive.
16	EXPERIMENTAL Design Experiments devise situations in which predicted outcomes may be evaluated by observations. Individuals on which action is taken, e.g. given a new drug, are called ‘treatment group’. The ‘control’ group evaluates what happens in the ‘no special treatment’ individuals. Testing the effect of a drug in people normally requires giving a ‘placebo’, which creates the illusion that a drug is being taken when no drug is actually in the ‘pill’.
17	Experimental Design What is the appropriate control? Usually there are a number of possibilities. Randomization of individuals among treatments is important to get a fair test. It is important to avoid temporal or spatial structure of treatments. Larger sample sizes (more individuals evaluated) lead to more precision of estimated effect.
18	Drug Effectiveness Treatments may change the probabilities of outcomes for members of a population. Consider evaluating the effectiveness of a drug in reducing cancer incidence. Both outcomes, cancer & no cancer, occur in individuals getting the drug and in the control. Statistical analysis of outcomes tell whether or not the probabilities of specific outcomes differ according to treatments. One statistic is the contingency chi-squared.
19	Medieval Science Lab One of the most important decisions is what hypothesis to test. In the week 3 lab, you test the hypothesis that mediation and a crystal and cure a mutant fly. It is a long ago settled issue, but is serves to illustrate how to do statistics. When you have two groups (treatment and control) and categorical outcomes, the appropriate statistic is contingency Chi-squared. Hypothesis predicts crystal higher % ‘cure’ than control.
20	Contingency Chi-squared
21	Problem Your hypothesis is that males prefer Barack Obama to Hilary Clinton and the reverse is true for females. You ask a sample of people in downtown Chicago to express a preference. The results are:
22	Vocabulary Control Treatment group Randomization Replication Reproducibility Repeatability Testable hypothesis Measurement variable Categorical variable Subpopulation Statistic Null hypothesis