Hypotheses
A hypothesis should:
- be feasible (i.e. be sensible and based on scientific concepts)
- be a statement (not a question)
- be based on observations
- involve one independent and one dependent variable in a cause-and-effect relationship
- be testable and measurable in a way that demonstrates cause and effect
Experimental variables
| “Cause” | INDEPENDENT VARIABLE (IV) the thing that the experimenter deliberately varies/manipulates | Controlled variables are those that are kept constant during an experiment because they might affect the DV. | |
| ↓ | |||
| “Effect” | DEPENDENT VARIABLE (DV): the thing that the experimenter measures |
Validity
The concept of Validity relates to whether the measurements you are taking are caused by the phenomena you are interested in. The concept of a “fair test” is closely related to the concept of validity.
A method is valid if:
- variables are controlled. Uncontrolled experiments are never valid because you can never be sure whether the measured changes in the DV are caused by the IV or by some other, uncontrolled variable.
- appropriate measuring equipment and procedures are included (e.g. measuring cylinders are used to measure volume rather than a beaker)
- it investigates what you think it will investigate (i.e. the procedure actually tests the hypothesis and the experiment includes an appropriate range of values). In other words, the measurements are actually measuring what you intend them to measure.
Conclusions are valid if the results of in an experiment are relevant to the conclusion.
Reliability
A reliable experiment has results which can be obtained consistently.
To assess whether the results of an experiment are reliable:
- the experiment must be repeated and consistent results obtained (within an acceptable margin of error) Note: Repetition will only allow reliability to be assessed (it will NOT improve it).
To design an experiment that will produce reliable results:
- Reliability can be improved by carefully controlling all variables other than the DV and IV.
- When variables are not controlled, experiments often produce unreliable results: different trials produce inconsistent results.
Reliability and validity are often affected together
A failure to control all variables can produce experiments that are both invalid and unreliable. When an experiment inadequately controls variables that could affect the DV, it is an invalid experiment. It will also produce unreliable results due to the uncontrolled variables. Unreliable experiments are never valid because you can never be sure whether the measurements you are taking are caused by the IV.
On the other hand, experiments can produce reliable results but be invalid. Measurements and other observations can be reliable without being valid. For example, a faulty measuring device can consistently provide a wrong value therefore providing reliably incorrect results.
Reliability and repetition
It is important to repeat an experiment in order to assess whether an experiment is both reliable and valid. If the repeated results of the experiment show it to be unreliable, then it must be because of some uncontrolled variables and the experiment is also invalid.
Repetition does not improve reliability. The results of an unreliable experiment will not become more consistent and reliable if the experiment is repeated unless the cause of the unreliability is addressed.
Accuracy
To assess whether the results of an experiment are accurate:
- Compare the results to information in reliable secondary sources (e.g. scientific reports). If the results are close to the accepted or ‘true’ value of the quantity being measured, then they are accurate.
To design an experiment that will produce accurate results:
- The experiment must be valid and the sensitivity of the measurement equipment must be appropriate.
Accuracy and repetition:
Biological experiments often involve:
- measuring a variable with natural variation (e.g. the mass of individual organisms),
- measuring the probability of an event occurring (e.g. in genetics experiments)
- random uncertainty (or random error) in measurements. Most experiments involve measuring things with some degree of error, called measurement error. Random measurement errors can be caused by unpredictable fluctuations in the readings of a measurement apparatus, or in the experimenter’s interpretation of the instrumental reading.
In all of these cases, the accuracy of the results will be improved by repetition. This is essentially “the law of large numbers”: the average of the results obtained from a large number of trials should be close to the expected value, and will tend to become closer as more trials are performed.
In summary: experiments should be repeated to:
- Allow the reliability of the results to be assessed (see above)
- Allow anomalous data or outliers to be detected, disregarded or removed (if appropriate).
- Minimise the effect of random errors (which improves accuracy).
- Improve the accuracy of measurements of variables with natural variation, or measurements of the probability of an event occurring
Advanced
additional information goes here