click below
click below
Normal Size Small Size show me how
Statistics
Statistics slides
Question | Answer |
---|---|
What are the assumptions for a dependent t-test? | Level of measurement is interval or ratio, the data is normally distributed, there was random sampling and/or assignment, there was equality of variance. |
What are the null and alternative hypotheses of a dependent t-test? | Null=there is no significant difference between means of sample group at time 1 and time 2 Alternative=there is a significant difference between sample group at time 1 and time 2. |
What is the research question behind dependent t-tests? | Is there a difference between group means at time 1 and time 2? |
What does the confidence interval for a dependent t-test tell us? | CI means that under repeated sampling 95% of CIs would contain the true population stat. |
What are the assumptions for an independent t-test? | Random sampling and/or assignment, independent observations, normally distributed, equality of variance, mutually exclusive IVs. |
What are the null and alternative hypotheses of an independent t-test? | Null=there was no difference between group means Alternative=there was a difference between group means |
What is the research Q behind independent t-tests? | Is there a difference in the means of two groups undergoing the same treatment? |
What does the confidence interval for an independent t-test tell us? | CI means that under repeated sampling 95% of CIs would contain the true population stat. |
What are the assumptions for an ANOVA? | Equality of variances, random sampling and assignment, normally distributed, independent observations. |
What is an effect size? (N & S p. 59, Pallant , 207-207 & 247) | The relative magnitude of the differences between means – if there’s a difference, is it an important one, basically. It is calculated by dividing the sum of the squares between-groups by the total sum of squares. |
What is the line equation (p. 134)? | Y = a+bx, where a is the intercept (the value of Y when x is equal to zero) and b is the slope. |
What is a definition of a regression line (p. 134)? | The straight line that passes through the data and minimizes the sum of the squared differences between the fitted and actual data points. |
Why is it called least squares regression (p. 134)? | Because it tries to find a line that has the least squared sum of differences between the actual and fitted data. |
What is a correlation coefficient (p. 137)? | A numerical, descriptive measure of the strength of the linear relationship between two variables. |
What is R2, the coefficient of determination? | The proportion of variance in the DV that is explained by the IV. |
How to interpret correlation coefficient? | The value between –1 and 1 that possesses the same negativity or positivity of the correlation (slope of the line) and has magnitude that expresses the degree of linear association between two variables. |
What is Pearson’s correlation? (p. 137) | Obtained by dividing the covariance of the two variables by the product of their standard deviations. It’s called r. |
Read a scatter plot and generate one as done on p. 128 | Graphs, Legacy Dialogues, Simple Scatter, Define. DV to the Y-axis, IV to the X-axis. Shows us outliers, checks for homoscedasticity, shows us what direction its correlated, if it is. |
How do you find ID number of outlier? | Data Label Mode icon on Chart Editor. Double click, click on bullseye and clock on outlier. |
What did Cohen say about guidelines for interpreting correlation coefficients (p. 132)? | Small=.10 to .29 Medium=.3 to .49 Large=.5 to 1. |
How to interpret a Pearson vs. Spearman correlation? | If there’s a negative sign in front of the correlation coefficient value, there’s a negative correlation. The further the correlation coefficient is from zero, the stronger the correlation – .1 to .29 is small, .3 to .49 is medium, and .5 to 1 is large. |
What assumptions a Spearman correlation makes. | It’s linear? |
What is a partial correlation? What research question does it address, p. 142-143 | Allows us to control for an additional variable, usually because you suspect it’s influencing other variables. Research Q=after controlling this variable, is there still a relationship between two others? |
How to read the printouts and run result for partial correlation (p. 144) | Analyze, correlate, partial. If the correlation coefficient varies between the two sections, the third variable does have an effect. |
What are the three types of regression? ( p. 147-148) | 1) Standard multiple regression, all IVs entered into the equation simultaneously. Heirarchical or sequential, IVs entered in blocks. Stepwise, lets SPSS take control. |
What are assumptions about regression? | Sufficient sample size, no multicollinearity or singularity, outliers dealt with, ND, linear, homoscedasticity, |
Correlation assumptions? | Interval or continuous level of measurement. Both pieces of information, X and Y, are from the same person (related pairs). Independence of observatiosn – observation 1 doesn’t influence observation 2. ND. Linear relationship. Homoscedasticity. |
Correlation research Q? | Is there a relationship between X and Y? |
Correlation null hypothesis? | There is no relationship. |
Correlation alternative hypothesis? | There is a relationship. |
What does Levene’s test tell us?? | Whether or not the variances in the groups are the same – whether or not the assumption of homogeneity has been violated. |
What does a post-hoc test tell us and why is this needed for an ANOVA and not for a t-test? | Used if the null hypothesis is rejected, to see which groups vary significantly. |
What is the definition of degrees of freedom (p.72) | Dfb (between groups) = number of groups minus 1. Dfw (within groups) = sample size minus number of groups. Dft (total) = Sample size minus one. |
What are two differences between a z-test and t-test? (p.71) | The t-test is used for two groups while the z-test is used for one. We cannot know the standard deviation for a t-test, whereas we know the standard deviation for a z-test. |
What are the rules about confidence intervals noted by Cunning and Finch on p. 74 of N&S? 1) | If the error bars don’t overlap, then the groups are sig at p=less than or equal to .01. 2) If amt of overlap=less than half of the CI, the sig. level=less than or equal to .05. 3) If overlap=more than half of the CI, dif is not statistically sig. |
What are the rules about Cohen’s effect size ( p. 74) | Effect size of .2 is considered to be small, .5 is moderate and .8 is large. |
What is a grand mean? | Mean of the means of several samples. |
What is the Sum of squares between groups? | Sum of the squared differences between the group means and the grand mean. |
What is the Sum of squares within? | Sum of squared differences between individual data and the group mean within each group. |
What is the Mean square between? | Sum of squares between divided by df. |
What is the Mean Square within? | Sum of squares within divided by df. |
What is a type 1 error? | Rejecting the null hypothesis when it is true. |
What is a type II error? | Accepting the null hypothesis when it is false. |
What is the difference between a planned vs. post-hoc test? | Post hoc tests:we do all tests - weaker power because, for ex, we calculate 0.05/6=.0083 in post hoc, as opposed to using planned comparison where we don't do all the tests and instead just use the p-value of .05, which has more power than .0083. |
What is a Bonferroni correction used for? | To avoid making an alpha error – count up total number of comparisons you’ll make (k), then divide .05 by k. Don’t use if there are more than 5 groups. It overcompensates. |
How to interpret correlation coefficient? | (p.138 |
What do correlations tell us? | Describe the relationship between two continuous variables, in terms of strength and direction. |
Read a scatter plot and generate one as done on p. 128 | Graphs, Legacy Dialogues, Simple Scatter, Define. DV to the Y-axis, IV to the X-axis. Shows us outliers, checks for homoscedasticity, shows us what direction its correlated, if it is. |
How do you find ID number of outlier? | Data Label Mode icon on Chart Editor. Double click, click on bullseye and clock on outlier. |
What did Cohen say about guidelines for interpreting correlation coefficients (p. 132)? | Small=.10 to .29 Medium=.3 to .49 Large=.5 to 1. |
How to interpret a Pearson vs. Spearman correlation? | If there’s a negative sign in front of the correlation coefficient value, there’s a negative correlation. The further the correlation coefficient is from zero, the stronger the correlation – .1 to .29 is small, .3 to .49 is medium, and .5 to 1 is large. |
What assumptions a Spearman correlation makes. | It’s linear? |
What is a partial correlation? What research question does it address, p. 142-143 | Allows us to control for an additional variable, usually because you suspect it’s influencing other variables. Research Q=after controlling this variable, is there still a relationship between two others? |
How to read the printouts and run result for partial correlation (p. 144) | Analyze, correlate, partial. If the correlation coefficient varies between the two sections, the third variable does have an effect. |
What are the three types of regression? ( p. 147-148) | 1) Standard multiple regression, all IVs entered into the equation simultaneously. Heirarchical or sequential, IVs entered in blocks. Stepwise, lets SPSS take control. |
What are assumptions about regression? | Sufficient sample size, no multicollinearity or singularity, outliers dealt with, ND, linear, homoscedasticity, |
Why are outliers a problem? | They can have a large effect on means and other parts of the test. |
What is multicollinearity and singularity? | Multicollinearity exists when the IVs are highly correlated. Singularity occurs when one IV is actually a combination of other IVs. |
What are interactions over time? (see N & S p. 94-96) | Lines that are not parallel indicate that there is an interaction. |
Correlation assumptions? | Interval or continuous level of measurement. Both pieces of information, X and Y, are from the same person (related pairs). Independence of observatiosn – observation 1 doesn’t influence observation 2. ND. Linear relationship. Homoscedasticity. |
Correlation research Q? | Is there a relationship between X and Y? |
Correlation null hypothesis? | There is no relationship. |
Correlation alternative hypothesis? | There is a relationship. |
What is the difference between Levene’s test of homogeneity of variance and Levene’s test of equality of variance? | There is no difference |
What does Leven’s test tell us?? | Whether or not the variances in the groups are the same – whether or not the assumption of homogeneity has been violated. |
Why would one use a robust ANOVA test like Welsch or Brown- Forsyth (start p. 246 Pallant)? | When the Levene test has a result lower than .05, telling us that the assumption of homogeneity has been violated. |
What does a post-hoc test tell us and why is this needed for an ANOVA and not for a t-test? | Used if the null hypothesis is rejected, to see which groups vary significantly. |
What is the difference between clinically significant and statistically significant? | Statistically significant difference doesn’t make any claims about the magnitude of the effect of the difference |
What is the definition of degrees of freedom (p.72) | Dfb (between groups) = number of groups minus 1. Dfw (within groups) = sample size minus number of groups. Dft (total) = Sample size minus one. |
What are two differences between a z-test and t-test? (p.71) | The t-test is used for two groups while the z-test is used for one. We cannot know the standard deviation for a t-test, whereas we know the standard deviation for a z-test. |
What are the rules about confidence intervals noted by Cunning and Finch on p. 74 of N&S? 1) | If the error bars don’t overlap, then the groups are sig at p=less than or equal to .01. 2) If amt of overlap=less than half of the CI, the sig. level=less than or equal to .05. 3) If overlap=more than half of the CI, dif is not statistically sig. |
What are the rules about Cohen’s effect size ( p. 74) | Effect size of .2 is considered to be small, .5 is moderate and .8 is large. |
Why don’t we run a lot of t-tests instead of doing an ANOVA? (P.77-78) | ANOVA more efficient, less noise. Doing many t-tests increases the chance of alpha error, have to maintain the integrity. We can also use Mean Square term as a better estimate of within-group variance. |
What is an eta square? (N & S p.87, Pallant p. 247) | Variable that shows us the strength of statistically strong relationships. It always yields a number between 0 and 1 and is interpreted as a proportion of the variance in the DV that can be attributed to the IV. |
What is the difference between a planned vs. post-hoc test? | Post hoc tests:we do all tests - weaker power because, for ex, we calculate 0.05/6=.0083 in post hoc, as opposed to using planned comparison where we don't do all the tests and instead just use the p-value of .05, which has more power than .0083. |
What did Cronbach recommend here for a reliability level ? | Cronbach alpha coefficient of a scale should be above .7 |
What is homoscedasticity of errors?; | If the errors have constant variance, the errors are called homoscedastic. |