When conducting experiments and in particular in your ISAs, you will need to be able to use and understand the following words.
To improve accuracy take an average of repeats of the experiment. This reduces the effect of random error but has no effect on systematic error.
When deciding on a scale, think of the values of coins. Make your scale go up in either 1, 2, 5, 10, 20, 50 or 100's.
It is also sensible to use the 'coin values' when dealing with numbers less than 1. For example going up in 0.20's could be a good scale.
Going up in 4's can also be a good scale as it is easy to plot the numbers in between divisions.
If your graph is less than half a page in size, then go up in 2 squares thereby doubling your scale. Start in bottom left corner of the page to give yourself the most space possible, when it comes to graphs  big is better.
Bar Chart
NOTE: Rarely used in Science
Used to display data where one variable is Categoric (or Ordered) and the other Continuous. This can normally be easily recognised as one set of word values and one set of numerical values.
There are gaps between the bars as the edges of the bars bare no relationship to the next bar. The bars must be of equal width.
 For example: Time taken for a metal bar to conduct heat along its length
When drawing a bar chart for an ISA there are four marks available:
 X axis
 The bars are of equal width and not touching with the graph being more than ½ a page
 Axis is labelled
This comes from the results table heading
 The bars are correctly labelled
 Y axis
 Scale is linear (equally spaced values) and means the graph is more than ½ a page
 Axis is labelled
This comes from the results table heading
 Units
This comes from the results table heading
 Bars correctly plotted to within a millimetre

Histogram
NOTE: Rarely used in Science
Used to display frequency (counts / tally) data between interval categories. Here there are no gaps between bars, because the left hand side of one bar is just short of the value of the right hand side of the next bar.
 For example: Number of leaves within a specific range of length.
Line Graph
Used to display continuous variables, ie. numeric values on each axis.
NOTE: Rarely used in Science
A scatter graph with a dottodot line. This is used when there is extreme confidence in the data and there isn't likely to be an overriding trend / relationship. Theses are nearly always time related.
 For example: SpeedTime graphs, Growth with respect to time, (Business Profits)
Pie Chart
NOTE: Rarely used in Science
Used to display the relative proportions of a total
 For example: The relative proportions of sources of background radiation
Scatter Graph
NOTE: Most commonly used graph in Science
Used to display data with two sets of continuous variables. This can be easily recognised as two sets of numerical values.
Points are plotted and a smooth 'Line of Best Fit' is drawn. This shows the general trend and averages out any errors in the data. The 'Line of best Fit' on a scatter graph is the representation of our best guess at the true values.
The line could be straight ie. linear, or equally correct is a curved line. It is a judgement to be made when looking at the data points. Because of this, there should be at least 5 data points to enable the trend to be correctly identified.
 For example: Weight against Mass, Temperature against Time or Count Rate against Time
The independent variable is plotted on the xaxis (the axis at the bottom of the graph). The dependant variable is plotted on the yaxis (the axis at the size of the graph)
Common shapes of lines are:
When drawing a scatter graph for an ISA there are four marks available:
 X axis
 Scale is linear (equally spaced values) and means the graph is more than ½ a page
 Axis is labelled
This comes from the results table heading
 Units
This comes from the results table heading
 Y axis
 Scale is linear (equally spaced values) and means the graph is more than ½ a page
 Axis is labelled
This comes from the results table heading
 Units
This comes from the results table heading
 Points correctly plotted to within a millimetre  You must use a sharp pencil for this
 There is a thin (use a sharp pencil), smooth 'Line of Best Fit'
draw the 'Line of Best Fit' with a ruler if the points are close to being a straight line
 NB: If there is no pattern to the data, such that a 'Line of Best Fit' can not be drawn, then write on the graph: 'There is no correlation in the results' to gain the mark.

Sketch Graph
A sketch graph is just the line of best fit on labeled axis, no scales are required. The line of best fit should depoced the general tred, both the direction and shape of line. If the numbers in the table have roughly the same interval, it is probably a striaght line relationship
Linear Scale
A linear scale is one where each centimetre on the axes equals the same amount. For example, if the scale goes up in 10s, for every one centimetre square, then the scale is linear.
Errors
All readings have errors in them, as scientists we need to understand them and reduce there effect. They cause readings to be different from the true value. There are 3 different types:
Random Error
Random error, also known as experimental error, affect each measurement differently. They are normally due to a variation in a control variables that we are unable to keep exactly the same. For example the room temperature during the experiment  it is approximately the same, but small variation always occur.
Random errors are represented on a scatter graph by the distance from the 'Line of Best Fit'. In the diagram below the green lines show the error. It is assumed that the error in measuring the independent variable is low compared to that of the dependant variable  hence the green lines are vertical. If the points are all close to the 'Line of Best Fit' then we can say that our results are good as the error is minimal.
Random errors may be detected and compensated for by taking a large number of readings.
 For example: Random errors may be caused by human error, a faulty technique in taking the measurements, or by faulty equipment.
Systematic Error
These cause readings to be spread about some value other than the true value; in other words, all the readings are shifted one way or the other way from the true value. Systematic errors occur to the same amount every time that you take a particular reading and therefore, if detected, can be corrected.
 For example: A systematic error occurs when using a wrongly calibrated instrument.
Zero Error
These are a type of systematic error. They are caused by measuring instruments that have a false zero.
 For example: A zero error occurs when the needle on an ammeter fails to return to zero when no current flows, or when a toppan balance shows a reading when there is nothing placed on the pan.
Anomaly
An anomaly is a point significantly out of place compared to the other results. This is normally due to a misreading or a mistake in the recording / plotting of results. These should be ignored both when calculating a mean drawing a 'Line of Best Fit' as they are clearly wrong  beyond the realms of random error.
It is common practice to circle any anomalies, both on the graph and in a table of results.
Parallax
Parallax error is caused when reading from a needle (analogue) meter, when not looking straight on at the meter.
High end analogue meters have a mirrored back to eliminate parallax error. When reading such a meter, make sure that the reflection of the needle is behind the needle itself  this means that you must be looking straight on to the meter.
Digital meters do not suffer from this effect as it does not matter which angle you look at the display, it will always show the same reading.
Conclusion
The Conclusion is the final part of the analysis. It states what the results show and should answer the aim. It should be backed up with evidence from your results and then explained using Scientific Theory.
Remember to use:
Point  Evidence  Explain

In an ISA there are normally 3 marks available for the conclusion. You need to be as detailed as you can.
A good start is to describe the direction of the 'Line of Best Fit' and then its shape.
Half of the marks are for backing up your claims with numerical data from your results as a quantitative statement.

How to write a relationship
Evaluation
The Evaluation describes how sure you are of your conclusion and attempts to explain where errors came from. Also included is the ways in which you could improve the experiment.
Preliminary Test
The first part of an experiment in which the equipment is used to decide things about the actual experiment. For example:
Range
The range of a variable is the difference between the largest and the smallest values for a variable.
It is better to have as large a range as is possible for the independent variable, as this normally gives the biggest change in the dependant variable. This makes identifying any patterns that much easier and reduces the effect of any errors.
When quoting a range give the actual values used.
Interval
The interval is the gap between the chosen values for the independent variable. This should be chosen in conjunction with the range to produce at least 5 different readings.
While in most investigations the interval is constant, it is only a rough guide to make sure that we have data points spaced throughout the range and therefore the graph.
We sometimes may vary the interval to concentrate on a particular point in the relationship. This is especially true when a we have a 'Ushaped' Line of Best Fit. We would normally be interested finding the point at which the line turned. Although this may form a secondary experiment.
Reliability
Reliability is a measure of how much you can trust your conclusion drawn from your results.
The results of an investigation may be considered reliable if the results can be repeated. If someone else can carry out your investigation and get the same results, then your results are more likely to be reliable.
One way of checking reliability is to compare your results with those of others. The reliability of data can be improved by carrying out repeat measurements, for lots of different independent variable values as this allows the identification of anomalies while taking a mean reduces the effects of random errors.
You may not need to do repeats if:
 A second measurement matches exactly the first one
 The first set of measurements lie very close to a line of best fit. To know this you would need to plot the first set measurements before packing away.
Repeatable
Results are said to be repeatable if you can get the same results when you redo your experiment.
Reproducible
Results are said to be reproducible if someone else can get the same results when they redo your experiment.
Validity
Data is only valid for use in coming to a conclusion if the measurements taken are affected by a single independent variable only. Data is not valid if for example a fair test is not carried out or there is observer bias.
 For example: In an investigation to find the effect on the rate of a reaction when the concentration of the acid is changed, it is important that concentration is the only independent variable. If, during the investigation, the temperature also increased as you increased the concentration, this would also have an effect on your results and the data would no longer be valid.
Sensitivity
The sensitivity of an instrument refers to the smallest change in a value that can be detected.
 For example, bathroom scales are not sensitive enough to detect the weekly changes in the mass of a baby, whereas scales used by a midwife are sensitive enough to permit a growth chart to be plotted.
Resolution
The resolution of a measuring instrument is the smallest measurement it can measure, ie the smallest division marked on the scale, or possibly half of a division.
You can tell if you have used an measuring division with appropriate resolution if the difference between readings is much greater (5 times bigger) than the resolution used.
Calibration
This involves fixing known points and then marking a scale on a measuring instrument, using these fixed points. It can also be done to an electronic sensor's output.
The sensor would need to be placed in a known condition, which was measured by another device. You would then record the known condition and the sensors output. Finally you would mark the known condition on a scale that the sensor's output will be measured on.
 For example: To mark the lines on a thermometer, put the thermometer in just melting ice  which we know is at 0°C. Then mark 0°C on the thermometer at the height of the fluid. Next put the thermometer in just boiling water, mark this as 100°C. Then draw 100 equally spaced marks in between, to represent each degree
Numerical
A numerical value is one that is a number.
Quantitative
Quantitative statements are those for which a numerical value representing is expressed. It is quantitative statements that scientist try to produce. It is how you should justify your conclusions, as the evidence part of:
Point  Evidence  Explain

Qualitative
Qualitative statements are used if there is something you can't put a number on to give it its meaning or value, and therefore descriptions are used instead. This makes them subjective and not as useful as Quantitative statements to a scientist.
Optimum
The Optimum value is the best value to use. It is used to describe the top of a peak or the bottom of a trough on a graph  depending on what is best:
Any value less than the optimum produces not as good a result and any value greater than the optimum produces not as good a result.