Paper 3 Q2 - Analysing Data

prompts to remind you of some useful facts about data

This is a left (or negative) skew. Most scores were high, but a small minority were much lower. This suggests that most people responded to the manipulation of the IV in a similar way, but a small minority responded in a different way (i.e. participant variables influenced the outcome).

This is a right (or positive) skew. Most scores were low, but a small minority were much higher. This suggests that most people responded to the manipulation of the IV in a similar way, but a small minority responded in a different way (i.e. participant variables influenced the outcome).

If the mean is significantly lower than the median, this suggests there is …

a left (negative) skew

If the mean is significantly higher than the median, this suggests there is …

a right (positive) skew

If the standard deviation is large but the IQR is not, this suggests …

… the presence of outliers which have effected the sd but not the IQR.

this data is normally distributed - there is no skew and most scores are similar.

If the mean and the median are similar, this suggests that …

…the data is normally distributed

this shows a left (negative skew) in the middle 50% of data

this shows a right (positive skew) in the middle 50% of data

this shows that there is no skew in the middle 50% of data but there are outliers at the extreme upper limit of the range. This would suggest the IQR will be low but standard deviation will be high.

a strong, positive correlation means that …

as scores for one covariable increased, they also increased for a second covariable.

a strong, negative correlation means that …

as scores for one covariable increased, scores decreased for a second covariable.

bidirectional ambiguity

we are unable to know if an increase in factor A is causing an increase in factor X or if an increase on factor X is causing an increase in factor A

A high IQR indicates that …

… the middle 50% of data was widely spread around the median, i.e. participants did not respond in a consistent way - different people did different things

… the middle 50% of data was tightly packed around the median, i.e. participants responded in a consistent way - performance was the same (or similar) for most participants