2.1.2 data handling/ maths skills

primary data

data from a firsthand source or account collected specifically for the research

secondary data

data from a secondhand source, not collected by the researcher and not specifically for the purposes of the study

quantitative data

numerical data, displayed with:

\-tables

\-scattergrams

\-bar charts

\-histograms

\-line graphs

qualitative data

data that cannot be counted or measured, but instead describes a characteristic

meta analysis

a form of data analysis that looks at large sets of data over a number of studies to find a common pattern or conclusion

effect size

the dependent variable of the meta analysis which provides an overall measure of the relationship between variables

meta analysis evaluation

strengths:

\-results can be generalised

\-produces more accurate and well-evidenced conclusions, based on large sets of data

weaknesses:

\-can often be victim to publication bias

content analysis

technique for systematically describing the results of a study or communication quantitatively

may involve studying behaviour through media, and can inform about the beliefs of a society

measures of central tendency

the mean, median and mode are all (blank), meaning averages

measures of dispersion

the range and standard deviation are (blank), used to measure the spread of the data

standard deviation

used to measure how the scores are distributed around the mean

a large score means that there is a lot of variation

uses exact values, providing a more precise score than the range

normal distribution

also called the ‘bell curve’, a symmetrical spread of data with most scores in the middle, and average located at the highest peak

skewed distribution

spread of frequency is not symmetrical, and is clustered at either end

the mode remains at the peak, but the mean is dragged to either the left or the right as it is affected by extreme values

nominal data

data that is in categories or groups

ordinal data

data that can be ranked in order and is based on an arbitrary scale (eg a test score, where the median is used as the measure of central tendency)

interval data

data measured on a fixed scale or unit, where the mean is used as the measure of central tendency

negative skewed distribution

data is distributed so that the mean is the lowest value - most people have scored high

positive skewed distribution

data is distributed so that the mean is the highest value - most people have scored low

mean

average of a set of all values, but can be skewed by extreme values

mode

most frequent value in a set, but not useful if there are many different values with the same frequency

median

the middle score when a set of values is ordered numerically, unaffected by extreme scores but not necessarily reflective of the whole set

type of data, experimental design, correlation

three factors that affect which statistical test to use

type 1 error

accepting the hypothesis when it is not true - a false positive

type 2 error

rejecting the hypothesis when it is actually true - a false negative

spearman’s rho

statistical test:

-correlation coefficient

-ordinal data

pearson’s r

statistical test:

-correlation coefficient

-interval data in normal distribution

-value will fall between +1 and -1, indicating pos or neg correlation

wilcoxon

statistical test

-related design

-ordinal data

sign test

statistical test:

-related design

-nominal data

related t-test

statistical test:

-related design

-interval data

unrelated t-test

statistical test:

-unrelated design

-interval data

mann-whitney

statistical test:

-unrelated design

-ordinal data

chi-squared

statistical test:

-tests the null hypothesis by comparing results with expected frequency

-nominal data, either unrelated or correlational design

parametric test

related and unrelated t-test and pearson’s r are examples of a (blank), a test done on interval data in normal distribution which is more robust than other tests (better able to detect significance), so is preferable

it is crude and does not give a numerical value for each participant

what is a limitation of using nominal data?