Upgrade to remove ads
Reminders from Second year
Terms in this set (58)
What happens to the SEM as the sample size increases?
How is variance calculated?
Sum of squares/d.f.
Which measure of central tendency is most sensitive to outliers?
What is psuedoreplication?
Where there is exaggeration of statistical significance when observations have been taken from the same individual and are therefore interdependent
Can you increase 'n' by taking more samples from the same individual?
No, this is an example of pseudoreplication
What is a poisson distribution?
A discrete frequency distribution which gives the probability of an event occurring in a fixed time and/or space
What happens to the S.D. if 'n' increases?
NOTHING. The SD is calculated as the best possible estimate of the SD of the overall population
What do confidence intervals define?
The boundary in which we can confidently expect the population mean to fall
What does a 95% CI imply?
That you can be 95% sure that the population mean falls within the confidence interval
There is a 1 in 20 chance that it does not overlap with the population mean
Which error bars are most commonly used in biomedical science?
Why are CI bars problematic in biomedical science
Overlap does not necessarily mean that there is no statistically significant difference between the groups - this is still a possibility
What does it mean if SEM bars overlap on a graph?
Means do not differ at
What does it mean if SEM bars do not overlap on a graph?
Means differ at P<0.05
What can be said about the inference made from all error bars?
More statistical testing is required to confirm a statistical deduction
What does P<0.05 mean with regards to the null hypothesis?
There is less than a 5% chance that the results have arisen due to chance if the null hypothesis is true
How is the t statistic calculated?
(Sample mean-population mean)/SEM
What are the disadvantages of a within-subject design?
Confounding likely - perhaps the order in which treatments were given or carry-over effects of the treatment
Is the homogeneity of variance assumption used in the paired t-test?
No, since each group should have a similar variance and SD
What is the F ratio?
The ratio of the group mean square: residual mean sqaure
If H0 is true, what is the value of F?
Close to 1
If H0 is false, what is the value of the F ratio likely to be?
Greater than 1 - i.e. the bigger the F ratio, the greater the statistical difference
If the data do not meet parametric assumptions for an independent one-way ANOVA, what test must you use?
How is statistical power affected by group sizes?
Equal numbers in each group increase power
Larger the sample size, the larger the power
When do post-hoc tests perform badly?
When there are unequal group sizes and unequal variance
What is the Wilcoxon Signed Rank test?
A non-parametric test used when data for the (i) paired t-test (ii) one sample t-test do not meet parametric assumptions
What do the assumptions of non-parametric assumptions still rely on?
What is the non-parametric equivalent of the unpaired t-test?
Why aren't non-parametric tests used all the time?
Because they are liable to increased type II error rates - they throw away a lot of information and are therefore likely to miss effects
What is the null hypothesis associated with a Chi-squared test?
There is no association between the factors
What is the alternative hypothesis associated with Chi-Squared test?
There is an association between the factors
What happens if the standard error of the estimate is large?
It reduces the confidence in the values and therefore limits the analysis
What does multicollinearity do to a standard error?
It means it becomes uncertain which variable is contributing the effect
What is the z-test for the proportion?
A test calculating a 95% CI for the proportion and seeing if it differs to a specified proportion/how much it overlaps
In what circumstances must you not simply divide a P value by 2 to get from a 2 tailed to a 1 tailed p value? Why?
Fischer's exact test
It has a non-symmetrical hypergeometric distribution
The Wilcoxon Signed Rank test is a non-parametric equivalent to which statistical tests?
1. One sample t-test
2. Paired t-test
What is the NP equivalent of the one-sample t-test?
Wilcoxon signed rank test
What hypothesis is being tested in the one sample t-test?
The null hypothesis that the mean difference between the pairs is 0
What is the paired test to test that 2 proportions are equal?
If you are unable to run a Fischer's exact test on data which do not meet the assumption for Chi Sq, what must you do?
Combine the cells in a logical way
If you have >2 groups in a test for equal proportions, what are the requirements to run a Chi Sq. test?
>80% of the cells must have an expected frequency over 5
If you have two groups, what is the requirement to run a Chi Sq. test?
All of the cells must have an expected frequency over 5
What test tests for a trend across categories?
Linear by linear association for the trend
Is the linear by linear association test independent or paired?
What is the only 'paired' test when testing for differences in proportions?
When must you use a Spearman's Rank correlation?
When one or both of the variables are not normally distributed
For data that are not normally distributed by do have a central tendency, what are the most informative summary statistics?
What is the overall type I error rate for a multiple t-test comparison of 5 groups?
Answer = 0.95*0.95*0.95
= 0.95*0.95*0.95*0.95*0.95 = 0.77
1 - 0.77 = 0.23
Type 1 error rate has increased from 5% to 23% - therefore don't do multiple t-tests for more than 2 groups!Wh
What is the risk of using a p value of p<0.10 rather than p<0.05?
A greater risk of a type I error
How is a 95% confidence interval calculated for the SEP?
95% CI for the proportion =
+/- 2 x SEP
Then +/- this to the proportion as calculated
True or false: P <0.001 indicates a bigger effect size than P <0.05?
p values do not have anything to do with effect size
True or false: P >0.05 means that the alternative hypothesis is true?
P values deal with probabilities not certainties
When should the mode be used as a measure of central tendency?
ONLY when the data are nominal
What are the d.f.'s associated when reporting a Pearson correlation coefficient?
What are the degrees of freedom associated with the linear-by-linear association test?
Since you have specified what score one of the proportions is, you have ranked the scores in a linear sequence and therefore there is only one d.f. for the data to vary
How can bias be avoided?
Researchers and participants unaware of intervention (double-blinded study)
How are controls most effectively used?
If they closely match the experimental intervention i.e. injection of saline in same way as injection of the drug has been injected/sham surgery
What happens to the SD if the sample size decreases?
Will likely increase since the influence of one data point will have a greater influence on the overall data set
Why can't you simply rely on Shapiro-Wilks test to test for normality?
The test lacks power, and even substantial departures from normality will not be significant
Sets found in the same folder
Statistics: From Pete's Hints and Tips
Sets with similar terms
Statistics for Data Analytics
10 - Inferential Statistics
Item Bank Worksheet Ch 9,10,11 stats- PY 211
Other sets by this creator
Using energy transfers from reactions
Making soap and biodiesel
Making NPK fertilisers