Data Analysis Flashcards
Give reasons why standard deviation is more useful than range when comparing variations
- not affected by outliers (anomalies)
- allows statistical tests to be calculated
Suggest why variables are controlled in an investigation
- observe effect of dependent variable alone to allow comparison
- control variables for validity
Suggest how to approximate highest and lowest value of a distribution given mean and standard deviation
mean plus/minus two times standard deviation
Suggest why control in drugs test is given existing treatment
unethical to give no treatment for (life threatening) health condition
Suggest what is given to a control group and why
- placebo given which similar to test in every way expect for having effect
- eliminate variables not due to treatment
Describe how scientists decide how many trials are appropriate
- test different numbers of trials
- select which ever gives best results in appropriate time / number of trials after which there is no change in sample collected
Gives reasons for carrying out repeats
- exclude anomalous results
- calculate more reliable mean
- allow statistics to be carried out
- check results are concordant
Factors to consider when drawing a conclusion on data
- standard deviation overlap
- whether statistic is calculated
- if statistic shows significant trend
- how highly significant trend is
- could other factors influence result
- will experimental model give same effects in reality
- compare conclusion to experiment
Evaluate mean as a measure of average
- takes into account all data so gives accurate representation
- distorted by outliers
Evaluate median as a measure of average
- less effected by outliers
- not suitable for small data sets with large variation
Evaluate mode as a measure of average
- unaffected by outliers
- may not give accurate representation of data set
- could be multiple or no mean
Tips for describing data
- increase/decrease
- gradually/ rapidly
- between ______ and _______
- highest/lowest at _______
- fluctuates/CONSIDERABLE variation which may be due to chance
- start and end same / different
- _______ higher/lower than _______ throughout
- follow similar pattern / describe pattern EXACTLY
- never say CHANGE - specify increase/decrease
- do not describe graph if asked what effect it shows (interpret!)
Describe how a statistical test would be used to find a significant correlation or difference (in means or frequencies)
- construct a null hypothesis
- carry out statistical test to calculate test statistic
- look up value in critical value table for p=0.05
- use figure to accept or reject null hypothesis (if test statistic is greater than critical value then reject null hypothesis)
CANNOT SAY USE CALCULATED VALUE TO ACCEPT/REJECT NULL HYPOTHESIS
Give a reason why student t test is the appropriate statistical test for this investigation
testing significant difference between TWO MEANS
Give a reason why Spearman’s rank is the appropriate statistical test for this investigation
testing significant correlation between two DIFFERENT CONTINUOUS variables
Give a reason why chi-squared is the appropriate statistical test for this investigation
testing significant difference between CATEGORICAL data
What do results from statistical tests suggest about correlation or differences in data being due to chance
p>0.05 means likely due to chance
p<0.05 means unlikely due to chance (p
Points to include in conclusion
- difference/ correlation significant or not
- reject or accept null hypothesis (even if not stated in question)
- contextual description of what this means
- (justification using statistical tests)
What does significant means
not due to chance
Suggest why standard deviation might be a better measure of variation in a population than range
- ranges could be similar
- suggests similar variation in populations
- standard deviation smaller in one population
- suggests more uniform distribution in one population compared to another
Points to include when evaluating a conclusion
- positive or negative correlation (draw line of best fit)
- correlation does not necessarily mean causation
- other factors could affect that have not been tested
- give evidence from data that opposes conclusion/ state variation in results
- statistical test not carried out / standard deviation bars overlap so difference could be due to chance
- be specific to conclusion (affect could be different in humans or only certain subjects tested which may not represent conclusion)
- repeats not been carried out
- limitations of control used
- additional limitations investigation (e.g. factors which may affect results that were not controlled unless told otherwise)
Evaluate whether stated value is most effective/highest from data
- not all values tested
- intervals tested are not small enough
- most effective/highest value could lie in this interval (state possible interval based on trend in data)
Describe how to plan an investigation to test a hypothesis
- state all factors which need to controlled and state how
- mention a suitable control if relevant
- mention a large enough sample needs to be taken
- describe how data will be collected
- determine if data collected is significant using a statistical test
Suggest why a large sample size is taken during an investigation
- representative
- minimise effect of anomalies
Suggest why samples are taken randomly during an investigation
avoid bias
Suggest why a scatter graph is suitable to plot
determine correlation between discrete/ independent variables
How to carry out an investigation to test for difference between populations
- random sample
- large enough to be representative
- measure
- calculate mean AND standard deviation
- student t test
- check for significance
Points to remember when doing a biological drawing
- sharp pencil
- scale bar
- ruled lines
- no shading / sketching
- title / heading
- label line should not overlap
- continuous lines
- state magnification
- correct proportions
When suggesting limitations to a method
- measurement taken standardised (possibility for bias/subjective??)
- are all necessary factors controlled/ standardised
- random selection of sample to avoid bias
