stats write ups Flashcards
partial correlation write up.
what do u include in note under correlation table ?
Note. Pearson’s product moment correlation r, with 95% confidence intervals around r. N = x. *p < .05, **p < .001
format for discussion of partial correlation
if decreases in strenght
while x increases with y, when z is controlled for, this relationship diminishes. this suggests that the relationship between x and y may be explained by z.
what to report in results section of partial correlations
- table with descriptive stats and product moment correlations with 95% CI around r.
- Report Pearson’s correlation significance in text with their direction if significant. (don’t need to report p value as they’re in table). e.g. pearson’s correlation coefficient revealed a significant positive relationship between x and y
- report partial correlation data. r(df) = .___, p < .___. report direction and change in significance
assumptions of regression
normality, linearity, (multicollinearity), homoscedasticity,
should u include correlations in regression model
only if multiple predictor variables
Results section for regression / multiple regression
- Only include correlations when more than one predictor variable
- Assumptions = normality, linearity, (multicollinearity), homoscedasticity,
- Test used = linear regression model
- Report F statistic AND R^2 value
- Clarification of F statistic significance
- If model is significant, report coefficient stats: b = _.__ [95% CI: , ]
- IF multiple predictors and result is significant, report t-tests for each.
- Direction should be clear.
- If multiple significant predictors, then say which one is a stronger predictor (based from regression coefficients)
Hierarchical regression results section
-descriptive stats table and correlations between study variables
- variables also including step construction: - E.g. the hierarchical regression was constructed to include workers salary in step 1, with their attendance in step 2.
- assumptions violated
Model 1, then change statistics in introducing 2. Then Final model with both.
e.g.
-salary alone (step 1) explained a significant proportion of variance in workers satisfaction, F(df, df) = ___, p < __ , R^2 = __ . Introducing attendance at stage 2 explained an additionl (R^2)% of the variance in satisfaction and change in R^2 was significant, F Δ(df, df) = __ , p = __.
-The final model, including both salary and attendance explained (R^2)% of satisfaction, F = …… .
if significant
- coefficient results b = [CI = ], t = , p = .
clarify which variable was strongest (if more than 1 signif)
95% of all sampled means fall within plus-or-minus1.96 SE of the population mean
true or false
true
marginal mean
factorial AVOVAS
the mean scores for each level of a single IV, averaged across all levels of the other IV
cell mean
factorial ANOVAS
the mean scores for each individual condition.
Bonferroni Corrected signif level
a/c c = no. of comparisons
Factorial ANOVA.
how to report post hoc for main effects
- If the main effect is significant and the IV has more than two levels, you should go on to report:
post hoc analyses using (Tukey HSD for between subjects, Bonferroni for within subjects). report significance (p value) and direction only .
compare each level of that IV.
Factorial ANOVA
how to report post hoc tests for interaction effect
in order to investigate interaction, we conducted __ (independent/paired) t-tests comparing the DV of the each level of the main IV across each secondary IV level.
Based on theses N pairwise tests, of simple effects, we applied the Bonferroni corrected criterion for significance of p < a/c.
These analyses showed that
Sort of combine all the effects in one sentence. E.g Use of secondary IV level A showed higher DV in main IV level 1 compared to main IV level 2, but not IV level 3.
Easier too look at graph when reporting.
report using this statistic
(t(df) = , p = , d = )
what are cell means
the mean scores for each individual condition (i.e. a single combination of levels from both IVs)
what are marginal means
the mean scores for each level of a single IV, averaged across all levels of the other IV.
e.g let’s say vindaloo scores have two levels for the other IV, red and white plate. You would take the average of those two. (indicating the mean score for that level.)
You could do the same for red plate score.
a significant main effect suggests that there is a significant difference between
the marginal means
(the scores for that IV)
posthoc analyses following a significant main effect examine the difference between the _____ means
marginal means (if more than 2 levels)
simple effects analyses following a significant interaction effect examine diff between ______ means
cell means
maulchys vs levenes test
maulchys tests sphericity. For repeated measures
Levene’s = test of homogeneity For independent groups
what test corrects df if sphericity is violated. what will happened to the df
Greenhouse-Geisser. the df will decrease
if Levene’s test is violated, what corrects for this?
use data from Welch.
df is adjusted to make the test more conservative. (almost like a penalty)
how to calculate df model and df error for repeated measures
k = number of iv levels in that iv
n = total number of p.p
dfm = k-1
dfe = (n - 1)(k-1)
how to calculate df model and df error for independent groups
k = no. of iv levels in IV
n = total no. of pp
dfm = k-1
dfe = n-k
repeated measures design with 5 IV levels and a total of 40 participants -
( __ , __ )
(4, 156)
an interaction indicates that the IV is having a significant effect on the DV, but that its effect is influenced by another IV
true or false
true
what is sphericity
homogeneity of covariance
i.e. the variance in difference scores under each IV level pair should be reasonably equivalent
