Quiz
EDS
This quiz has question types that open in the usual player.
Questions in this quiz (75)
1"Used to determine the dependence of a dependent variable on a predictor/independent variable". What test is this?
- ASimple linear regression
- BTwo-way ANOVA
- CChi-Squared test
- DPearson's test
- Erepeated measures ANOVA
2Assumptions of simple linear regression:
- ARandom sampling (allocation)
- BIndependence of data points (no pseudoreplication)
- CNormality
- DEquality of variance (homoscedasticity)
- EThere is a linear relationship
- FIndependence of residuals
- GNo entanglement
- HSphericity
3Which model diagnostics do you use to identify outliers?
- AStandardised residual greater than 2 is worth checking. If more than 5% of redisuals are greater than 1 means the model may be a poor fit
- BStandardised residual greater than 3 is worth checking. If more than 5% of redisuals are greater than 2 means the model may be a poor fit
- CStandardised residual greater than 4 is worth checking. If more than 5% of redisuals are greater than 3 means the model may be a poor fit
- DCook's distance of greater than 1 is a cause for concern
- ECook's distance of greater than 1.5 is a cause for concern
- FCook's distance of greater than 2 is a cause for concern
- GLeverage value 2(k+1)/n - anything over 0.2 may be exerting an effect
- HLeverage value (k+1)/n - anything over 0.4 may be exerting an effect
- IStandardised DFFit and standardised DFBeta - over 1 shows substantial influence on the model
- JStandardised DFFit and standardised DFBeta - over 2 shows substantial influence on the model
4"Compares 2+ independent variables against the dependent variable". What test is this?
- ASimple linear regression
- BTwo-way ANOVA
- CMixed model ANOVA
- DMultiple Linear Regression
- EGames-Howell Post-Hoc
5Assumptions of multiple linear regression
- ALinear relationship
- BEquality of variance (homoscedasticity)
- CIndependence of residuals
- DNormality
- ERandom sampling
- FNo entanglement
- GNo interactions
- HSphericity
6Multicolinearity is the [blank_start]strong[blank_end] correlation among predictors, or a [blank_start]weaker[blank_end] correlation of one predictor with multiple predictors. As colinearity [blank_start]increases[blank_end], the start errors of beta coefficients [blank_start]increases[blank_end] (so your degree of confidence in the slope coefficient [blank_start]decreases[blank_end]).
- Astrong
- Bweaker
- Cincreases
- Dincreases
- Edecreases
7How do you detect multicolinearity?
- AVariance inflation factor (VIF) of >10 for a variable is likely to be collinear with one or more coefficients.
- BAverage VIF >1 is problematic (overall there is a lot of multicollinearity)
- CTolerance factor (reciprocal of VIF/1 over VIF) <0.1 is problematic
- DTolerance factor (reciprocal of VIF/1 over VIF) <0.5 is problematic
- ETolerance factor (reciprocal of VIF/1 over VIF) <1 is problematic
- FAverage VIF >10 is problematic (overall there is a lot of multicollinearity)
- GAverage VIF >0.11 is problematic (overall there is a lot of multicollinearity)
- HVariance inflation factor (VIF) of >0.1 for a variable is likely to be collinear with one or more coefficients.
- IVariance inflation factor (VIF) of >1 for a variable is likely to be collinear with one or more coefficients.
8What can you do about multicolinearity?
- ACombine predictors e.g. weight and height into BMI. It should make biological sense
- BThere is nothing you can do. Soz.
- CTake one of the co-linear variables out - need to have a good reason to take it out.
- DUse a post-hoc test that can remove some of the colinearity and reduce VIF
9Which method is best to enter the predictors for a multiple linear regression model?
- AStepwise forwards
- BStepwise backwards
- CHierachical
- DAutomatic variable selection
10What is an unpaired t-test the same as?
- ASimple linear regression
- BOne-way ANOVA
- CTwo-way ANOVA
- DChi-squared test
- EPearson's correlation coefficient
- FMultiple linear regression
11What is an ANOVA the same as?
- ASimple linear regression
- BSimple linear regression for categorical variables
- CMultiple regression for categorical variables using dummy variables
- DMultiple regression of 2+ continuous variables
- EChi-squared test
12"This test analyses how much of the overall variance in the dataset can be explained by variation between group means (model) compared to the unexplained variation within a group (residual) i.e. tests the differences in means of 2 or more data sets". What test is it?
- AOne-way ANOVA
- BTwo-way ANOVA
- COne-way repeated measures ANOVA
- DTwo-way repeated measures ANOVA
- EMixed model ANOVA
13The F-ratio is the difference "between group MS"/"within group MS (residual". This is the signal:noise ratio – the [blank_start]higher[blank_end] the ratio, the higher signal to noise, so the [blank_start]more[blank_end] confident we can be that an affect is there. The ratio will be [blank_start]higher[blank_end] for the larger the difference in group means compared to the overall means, and the [blank_start]smaller[blank_end] the amount of residual variability. The P value associated with it says how likely an f-ratio this big or bigger is likely to occur by chance.
- Ahigher
- Blower
- Cmore
- Dless
- Ehigher
- Flower
- Gsmaller
- Hbigger
14General ANOVA assumptions:
- ARandom sampling (allocation)
- BIndependence of data points
- CNormality of the residuals
- DEquality of variance among groups (homoscedasticity)
- EThere is a linear relationship
- FNo entanglement
15The Levene's test assesses the equality of variance among groups. If the test is not significant, variances are equal.
- ATrue
- BFalse
16Which statements about post-hoc tests are true?
- APost hoc tests can be used to perform multiple comparisons whilst controlling overall type I error rate.
- BThey should be used when there is no specific hypothesis to be tested.
- CThey tell you which means are different after performing an ANOVA
- DYou can perform them before performing the ANOVA.
- EYou decide which to used based on the hypothesis you want to test.
- FAll post-hoc tests need equal sample size.
17Which statements about planned contrasts are true:
- AAre more powerful than post hoc tests.
- BAre less powerful than post hoc tests.
- CControl type 1 error rate better than post hoc tests.
- DControl type 2 error rate better than post hoc tests.
- EUsed to test a specific hypothesis.
- FYou should base it on a hypothesis made after the data is collected.
- GYou should base it on a hypothesis made before the data is collected.
18"Best to use if sample sizes and variance are equal, as it has a good trade-off between type I and type II errors." Which post hoc test is this?
- AGabriel's
- BTukey
- CGames-Howell
- DBonferronni
- EHockberg's GT2
- FDunnett's
19"This test corrects the p value by the number of tests being performed. It is very conservative (best to use when you need a high level of confidence in the results in any situation as is unlikely to make a false positive). Good control of type 1 error rate but poor control of type 2 error rate". Which post hoc test is this?
- ATukey
- BBonferronni
- CDunnett's
- DGabriel's
- EHocberg's GT2
- FGames-Howell
20"Best to use when you’re only interested in comparing all groups vs a single control group (also enables 1-tailed tests)." Which post hoc test is this?
- ATukey
- BBonferronni
- CDunnett's
- DGabriel's
- EHochber's GT2
- FGames-Howell
21"Use this test if sample sizes are slightly different". Which post hoc test is this?
- ATukey
- BBonferroni
- CDunnett's
- DGabriel's
- EHochberg's GT2
- FGames-Howell
22"Use this test if samples sizes are very different." Which post hoc test is this?
- ATukey
- BBonferronni
- CDunnett's
- DGabriel's
- EHochberg's GT2
- FGames-Howell
23If there is any doubt about the equality of variance assumption, then best to use this test. This compares every group with every other group". Which post-hoc test is this?
- ATukey
- BBonferronni
- CDunnett's
- DGabriel's
- EHochberg's GT2
- FGames-Howell
24Which statements are about orthogonal contrasts are true:
- ADon't inflate type 2 error rate.
- BDon't inflate type 1 error rate.
- CUsed to test a specific hypothesis.
- DUsed when there is no specific hypothesis to be tested.
- EThe two things we want to compare need to have opposite signs
- FYou must compare like for like - if you wanted to compare obese to all three groups, you need to multiple obese by 3.
- GThe order of the groups does not matter as long as they were sampled randomly.
- HThe sum of weights within a contrast must equal zero.
- IAssumes the data has no entanglement and that there is the independence of residuals.
25Which statements about polynomial contrasts are true:
- AThey are a type of orthogonal contrast
- BThey test for trends in the means.
- CThe order that the groups are entered in does not matter.
- DYou can only get linear trends, not quartic trends.
26"This statistic test has a two experimental factors being manipulated e.g. drug dose and mouse strain, which helps you see the interaction between two factors and how it impacts on the effect". What statistical test is this?
- AIndependent factorial ANOVA/Two-way ANOVA
- BOne-way ANOVA
- CMultiple linear regression
- DRepeated measures ANOVA
- EChi-squared test
- FMIxed model ANOVA
27Assumptions for a two-way ANOVA/independent factorial ANOVA?
- ARandom sampling
- BIndependence of data points
- CNormality of the residuals
- DEquality of variance
- ENo entanglement
- FSphericity
- GThere is a linear relationship
28How do you assess normality?
- AQ-Q plots
- BHistograms
- CShapiro-Wilk
- DLevene's test
- EKruskal-Wallis test
29How do you assess equality of variance?
- ADunnett's test
- BLevene's test
- CShapiro-Wilk
- DMauchy's test
30Which statements about the independent factorial/two-way ANOVA are true:
- ATwo experimental factors are manipulated e.g. sex and strain, so you can see the interaction between two factors and how it impacts the effect
- BIt has factors e.g. sex and levels e.g. male and female.
- CKruskal-Wallis test is the non-parametric equivalent.
- DOutputs are interaction plots.
- EIt produces an R square value.
- FYou can't use standard contrasts and post hoc tests if there is no main interaction found.
- GAt least one factor must be categorical.
- HEach level must have at least 3 levels.
31Interaction plots are produced from two-way ANOVA/independent factorial ANOVA. Which statements are true:
- AParallel lines indicate no significant interaction, but there still could be a main effect.
- BMain effect ignores the levels of one factor.
- CThe interaction looks at the levels of every factor.
- DIf there is an main effect, then interaction is not important.
- EYou can only test main effects with standard contrasts, not post hocs.
32"For this test, you in effect extend a paired t-test for 2 or more conditions. It looks at differences within the subjects and differences between subjects". What is this test?
- ARepeated measures ANOVA
- BPaired t-test
- CMultiple linear regression
- DMixed model ANOVA
- EOne-way ANOVA
33Which test is the paired t-test similar to?
- ARepeated measures ANOVA
- BChi-squared test
- CSimple linear regression
- DMultiple linear regression
- ETwo-way ANOVA
34What are the assumptions of a repeated measures ANOVA? (one-way or two-way)
- ARandom sampling (allocation)
- BIndependence of data points
- CNormality of residuals
- DEquality of variance among groups
- ESphericity
- FNo entanglement
- GThere is a linear relationship
35What test is used to assess sphericity?
- AMauchy's test
- BLevene's test
- CGabriel's test
- DEmma's test
- EShapiro-Wilk test
36Which statements about the two-way/factorial repeated measures ANOVA are true:
- AThere is more than one factor and they are repeated measures
- BTests for interactions.
- CDoesn't test for sphericty, like the one-way repeated measures ANOVA does.
- DThere is no non-parametric equivalent.
- ESidak is more reliable than Bonferronni for comparisons.
37Which statements about a mixed model ANOVA are true:
- ASphericity is not an assumption.
- BIt has the same assumptions essentially as any other factorial ANOVA.
- CThere is a factorial design with at least one independent factor and at least one repeated measures factor
- DThere is at least one categorical variable and one independent factor
- EYou can use simple effects post hoc analysis and pre-planned contrasts to find comparisons
- FUse bonferronni for repeated measures comparisons
- GUsed Sidak for repeated measures comparisions
38Which statements about simple effects analysis are true:
- AFinds where the interaction is happening after you have done the ANOVA
- BCompares all levels of one factor with all levels of the other factor
- CUse the LSD correction to control type 1 error rate.
- DIt takes into account the interaction between variables
39What correction do you use for simple effects analysis for a repeated measures ANOVA?
- ABonferronni
- BSidak
- CLSD
- DDunnet's
- EGabriel
- FGames-Howell
40What correction do you use for simple effects analysis for an independent factorial ANOVA?
- ASidak
- BBonferronni
- CLSD
- DGames-Howell
- EGabriel's
- FDunnett's
41When looking at a scatter plot of the standardised residuals for data that is in a model, what does this show you?
- AHomoscedasticity of the data
- BHow well the data fits the model
- CNormality of the residuals
- DWhether the model is two-way or one-way
- EIf the relationship is linear
42Which statements about multiple linear regression are true:
- Ayou are changing multiple independent variables to see how your dependent variable
- BYou are likely to get a better fit of the model to the data, because you are using more info to predict your dependent variable
- CSimple linear regression will have a higher R squared value because there is no colinearity
- DMulticolinearity could occur, but only if there are more than 3 predcitors
- EOnly remove a maximum of 2 outliers from each predictor data set (so if you had three perdictors, you could remove 6 data points)
43Which statements about R squared are true:
- AHow well the model fits the data
- BCalculated by SSm/SSt
- CThe value gives you the amount of variability within your data set that can be explained by the model you are trying to fit
- DThe bigger the R squared value, the higher the P value
- EIt tells you the size of the effect
- FAn R squared value of -77% means the model fits better than a model with an R squared value of 45%
44Which statements about the beta coefficient for regression are true:
- AThe steeper slope, the bigger the size of the effect
- BStandardised DFBeta is a standardised measures of how much the fit and slope of the model would change if the data point was removed
- CA beta coefficient of -3.4 means there is a bigger effect than a beta coefficient of 3
- DThe beta coefficient values won't change when you add more predictors
- EThe bigger the beta coefficient, the less chance that there is colinearity
- Fbeta coefficients are not reliable if you remove outliers
45Which correction can you use for ANOVA is sample sizes are unequal?
- ABonferronni
- BWelch's
- CLSD
- DSidak
- EGabriel's
46Why is performing multiple unpaired t-tests bad instead of using an ANOVA?
- AIncreases type 1 error rate
- BIncreases type 2 error rate
- CThe P value is less likely to be significant
- DYou can't use a post hoc on more than one t-test
47What is the problem with perfoming multiple post hoc tests?
- AIncreases type 1 error rate
- BIncreases type 2 error rate
- CThere is no issue
- DSPSS only lets you pick 2
- EIt increases the chances of outliers
4812 male C57BL/6 mice and 12 female C57BL/6 inbred mice are randomly allocated to 3 groups of drug treatments, with 4 mice from each sex in each group. There are 2 [blank_start]factors[blank_end] - sex and drug. Sex has [blank_start]2[blank_end] [blank_start]levels[blank_end]. Drug had [blank_start]3[blank_end] [blank_start]levels[blank_end].
- Afactors
- Blevels
- C2
- D1
- E3
- Flevels
- Gfactors
- Hgroups
- I3
- J2
- K1
- L4
- Mlevels
- Nfactors
- Ogroups
49Which tests require the sphericity assumption?
- ARepeated measures ANOVA
- BMixed model ANOVA
- CPaired t-test
- DRegression
- EOne-way ANOVA
- FTwo-way ANOVA
- GChi-sqaured test
50For polynomial regression, which is it bad to add lots of further terms e.g. go up to x^15?
- AIt over fits the model so doesn't generalise well to real world situations
- BIt risks decreasing the R squared value as it picks up fewer random fluctuations
- CIt makes the data linear when the data isn't actually linear
- DIt indirectly transforms the data so changes the relationship
51Which statements are true for non-linear regression:
- ANon-linear regression looks at the relationship between the X and Y variable (based on theory) in terms of various different parameters
- BIt transforms the data to improve the R squared
- CIt is equivalent to polynomial regression
- DIf it finds the local minimum and not the global minimum, it will not give you the best fit of the line, so you may need to change the starting parameters
- EThe computer guesses the starting parameters to predict the equation
52When do you constrain data for non-linear regression?
- AWhen you completely understanding the biology
- BOnly use it for constraining drug doses
- CWhen you need to improve the R squared
- DDon’t constrain unless there is a goof theoretical reason to do so, as it may worsen the fit
- EYou constrain so you can't get a negative minimum
53Assumptions for non-linear regression:
- ARandom sampling (allocation)
- BIndependence of data points (no pseudoreplication)
- CNormality of sampling distribution/normal distribution of residuals
- DEquality of variance among group (homoscedasticity)
- ENo entanglement
- FSphericity
- GThere is a linear relationship
54Which statements are true for a chi-squared test for two categories:
- AExpected frequencies in all cells should not be lower than 1.
- BExpected frequencies in al cells should not be lower than 5.
- CExpected frequencies in all cells should not be lower than 10.
- DIt compares two proportions
- EIt compares percentages
- FIf Chi-squared test is significant, you can reject the null hypothesis and accept the alternative hypothesis, that there is an association between the factors
- GIf Chi-squared test is significant, you can accept the null hypothesis and reject the alternative hypothesis, that there is no association between the factors
55Which statements are true for a chi-squared test for more than 2 categories:
- AAt least 80% of the expected frequencies must be equal to or lower than 5
- BAt least 90% of the expected frequencies must be equal to or lower than 5
- CAt least 70% of the expected frequencies must be equal to or lower than 5
- DIf the data is ordered, you can perform post hoc tests
- EIf the data is ordered, you can perform a linear association trend
- FIf the data is ordered, you can perform multiple comparisons
56What test do you use if the assumptions for the Chi-squared test are not met?
- AFischer's exact test
- BKruskal-Wallis test
- CSidak correction
- DFriedman's test
- EThere is no non-parametric alternative
57What is the non-parametric alternative for a factorial/two-way ANOVA?
- AThere is no non-parametric alternative
- BFriedman's
- CKruskal-Wallis
- DWelch's
- EFischer's exact
58What is the non-parametric alternative for a mixed model ANOVA?
- AThere is no non-parametric alternative
- BFischer's exact
- CKruskal-Wallis
- DFriedman
- EGabriel's
59What does the "Chi-squared test for trend" do?
- AThis test allows you to ask a more specific hypothesis to see if there is a linear trend.
- BIt tests for associations between new factors added into the contingency table after you have perform the Chi-squared test
- CIt is used to make the result more generalised to the population, so is similar to the adjusted R sqaured
- DIt helps corrected the raised type 2 error rate as a result of the Chi-squared analysis
60Power is the probability that you will be able to detect a difference if one is really there to be detected.
Power = 1 – [blank_start]type II[blank_end] error rate
Power is increased by:
o [blank_start]Increase[blank_end] effect size
o [blank_start]Decrease[blank_end] random variation
o [blank_start]Increase[blank_end] sample size
• If you increase the power, it is [blank_start]easier[blank_end] to detect an effect
• Having a [blank_start]low[blank_end] power increases the type 2 error rate, so you are [blank_start]more[blank_end] likely to accept the null hypothesis when there really is a difference among groups- Atype II
- Btype I
- CIncrease
- DDecrease
- EDecrease
- FIncrease
- GIncrease
- HDecrease
- Ieasier
- Jharder
- Klow
- Lhigh
- Mmore
- Nless
61o 68%: mean ± [blank_start]1[blank_end] x standard error
o 95%: mean ± [blank_start]1.96[blank_end] x standard error
o 99%: mean ±[blank_start]2.58[blank_end] x standard error
o 99.9%: mean ±[blank_start]3.29[blank_end] x standard error- A1
- B2
- C2.6
- D3.3
62Calculate confidence intervals
o 95% CI = ± [blank_start]2.0[blank_end] x SEM
o 99% CI = ±[blank_start]2.6[blank_end] x SEM
o 99.9% CI = ±[blank_start]3.3[blank_end] x SEM- A2.0
- B2.6
- C3.3
63Type 1 error rate: false [blank_start]positive[blank_end] e.g. drug [blank_start]doesn’t[blank_end] actually affect the variable in real life, but [blank_start]an effect[blank_end] is seen so the null hypothesis is [blank_start]rejected[blank_end].
Typw 2 error rate: false [blank_start]negative[blank_end].e.g. drug [blank_start]does[blank_end] affect the variable in real life, but [blank_start]no effect[blank_end] is seen, so the null hypothesis is [blank_start]accepted[blank_end].- Apositive
- Bnegative
- Cdoesn’t
- Ddoes
- Ean effect
- Fno effect
- Grejected
- Haccepted
- Inegative
- Jpositive
- Kdoes
- Ldoesn't
- Mno effect
- Nan effect
- Oaccepted
- Prejected
64What test is used to assess the normality of data points?
- AShapiro-Wilk
- BLevene's test
- CFriedman's
- DWelch's
65Skewness:
The data peaks to the right, so the data is [blank_start]negatively[blank_end] skewed. Transform this day with a [blank_start]square or cube[blank_end].
The data peaks to the left, so the data is [blank_start]positively[blank_end] skewed. Transform the data with a [blank_start]log10 or square root[blank_end].- Anegatively
- Bpositively
- Csquare or cube
- Dlog10 or square root
- Epositively
- Fnegatively
- Glog10 or square root
- Hsquare or cube
66What is the non-parametric alternative for a one-way independent ANOVA?
- AKruskal-Wallis
- BFriedman's
- CFischer's exact
- DGabriel's
- ESpearman's rank
67What is the non-parametric alternative for a one-way repeated measures ANOVA?
- AFriedman's
- BKruskal-Wallis
- CSpearman's Rank
- DFischer's exact
- EGames-Howell
681. For data that are not normally distributed but do have a central tendency, the most informative summary statistic would be the median and +- what?
- AInterquartile range
- BMean
- CMedian
- DMode
- ERange
- FStandard deviation
69Normally distributed variables X and Y are significantly correlated with a p level of 0.006 and a Pearson’s correlation coefficient of 0.325. Approximately how much of the variability in X and Y can be explained by this correlation?
- A23%
- B33%
- C11%
- D32%
- E1%
70What is the likely outcome of testing a null hypothesis using a 10% (i.e. P<0.10) level of significance rather than 5%?
- AIncreases type 1 error rate
- BIncreases type 2 error rate
- CDecreases type 1 error rate
- DDecreases type 2 error rate
714. A sample of 36 inbred mice had a mean weight of 25g and a SD of 3g. how large a sample of outbred mice having a mean weight of 30g and a SD of 5g would be required in order that the samples could have a similar 99% confidence interval for the mean?
- A100
- B75
- C25
- D60
- E72
72A skin infection was successfully treated in 40 out of 60 patients treated with a standard antifungal cream, compared to successful treatment of 45 out of 48 patients with an antifungal cream containing steroids.
What is the approximate 95% confidence interval for the percentage of patients successfully treated with the standard antifungal cream?- A54% and 79%
- B64% and 82%
- C43% and 65%
- D57% and 62%
- E67% and 81%
73What would be the outcome of increasing the sample size to 1000?
- AIncreases the significance of R squared
- BIncreases the P value
- CIncreases the size of the beta coefficient
- DIcreases type 1 error rate
74You can calculate residual variance by: 1 - R squared
- ATrue
- BFalse
75What summary statistics should you use with bimodal data?
- AThere is no best single summary statistic
- BMean
- CMedian
- DMode
- ERange
Information
- Description
- Experimental Design and Statistics (EDS) Quiz on EDS, created by Emma Jackson on 02/01/2017.
- Author
- Emma Jackson
- Language
- English
- Views
- 33
- Tags
- Rating
- No ratings yet Rate this quiz