Flashcards
part 2 stats
Flashcards do conjunto (45)
- 1Assumptions of linear model1) linearity and additivity
2) spherical errors
- independence of errors
- homoscedasticity
3) normality
- residual
- distribution - 2assumptions of linear model : name 2 spherical errors and what they are1) homoscedasticity: variance in residuals should be constant (data in scatter plot evenly spread)
2) independent errors: error in prediction for one set of data should be independent (not influence) the other - 3Assumptions of linear model: normality name them and what they are- normality of residuals: get graph on SPSS and see if the data follows a linear line very tight to line best fit - this is normality of residuals
- normal sample distribution: bell curve - 4homoscedasticty VS heteroscedasticity look like
- 5what is central limit theoremas long as you have a large sample you will get a normal distribution
- 6what is a bootstraprobust procedure on SPSS we use if we have a small sample (below 30) to get a normal distribution
- this will take many samples from the data and put each case back to maybe used again - 7how to detect for outliers- graph
- standardised residual
- cooks distance - 8cooks distanceway of detecting outliers, and above 1 cause concern - seen in SPSS data as 'coo_1'
- 9standardised residuals (detecting outliers)look at case wise output (STD. Resid)
- any values above 3 are outliers
95% scores should be +2
99% should be +2.5
any above 3 are outliers - 10how to work out a question like 'what % of standardised residuals above 2.5'look at case wise - standardised residuals
- count how many scores larger than 2.5 (include even if it has a miss infant of it)
(number of scores / n) X p = the % - 11f statistic and how to work it outratio of how well the model fits to how much variance
f = Msm / Msr
msm = averrage variability explained by model
msr = how much variability not explained for by the model - 12total sum of squares (SST)total variability
- 13model sum of squares (SSm / Msm)variability explained for by the model
- 14residual sum of squares (SSr/ Msr)how much variance is not explained
- 15Cohens d (small, medium, large)effect size, large good
0.2 = small
0.5 = medium
0.8 = large - 16simple cohens d equation is (mean 1 - mean2) / SD how do we know which means SD to use- the control group
- if both groups have equal variance, assume theyre same and pick either
- do a pooled estimate (the large equation) - 17post hoc?compare predictors when no hypothesis and not linear
- 18dummy and contrast codingcompare categorical predictors
dummy: only compares two things
contrast: compares more - 19problems with post hoc and how to fixincrease likelihood of type 1 error - saying it false when it is true
- fix by adjusting alpha with a diff post hoc such as Bonferroni - 20post hoc- use when no hypothesis
- same as contrast and dummy but linear
- compares all groups independently - 21covariateany variable that can impact outcome, that we adjust for
- 22if covariate has sig impact on results what does this mean for homogeneitymeans it has been broken
- 23how to report a covariatesame as normal f
f(df, df error) = f, p = - 24for covariates to know if it is sig or not from looking at graph, graph should look like?...if graphs look different then there was a sig impact of covariate, if graphs look same then no sig impact of covariate (good)
- 25if get sig finding of covariate what does this meanit means the covariate has impacted the results which is bad, which means homogeneity of variance has been broken
- 26why do we do covariatewhen we think an extraneous variable may be effecting results, so used to adjust means for EV / confounding
- 27advantages of covariatefiguring out covariates means more of th variance in the results is explained
allows us to adjust for them - 28in SPSS output - test within subjects showing covariate what does this table do to main effect if covariate sigif covariate is significant SPSS adjusts the main effect to adjust for th covariate interfering, so the main effect is the correct adjusted version
- 29contrast graph what does it mean if lines are parallel to each other and not parallelparallel lines means data not sig (as no sig difference in lines)
not equal lines = significant data (as sig difference between the lines) - 30how to work out odds and what if you get an odds answer over 1n event occurred / n even didn't occur
answers above one mean event occurred more than didn't and lower than one means event didn't occur more than it did - 31for log odds, the 'variable in the equation' output - what is B and what is exp(B) - what is constantB is expressed as logs
Exp(B) is as odds
- whatever you coded 0 as - 32for los odds do you read column B or Exp(B)Exp(B)
- 33explain b1 and b0 for log oddsyou code predictors as 1(b1) and 0 (b0)
- b1 means for every one unit increase you get an increase of...
- but for logs 1 unit change means changing from predictor 0 to predictor 1, so when changing from pudding to wine there was a (report b1 - if it is less than bo then..) decrease in deliveries (as b1 is lower than bo in output) - 34what is odds ratiochange in odds from one thing to another
elf example: change in odds (things getting delivered) from pudding to wine
- we saw a decrease in deliveries - 35what % of cases is allowed to be greater than 1.96 for standardised residuals5%
- 36assumptions of categorical outcomes- linearity of the logit
- spherical residuals
- independent errors
- complete separation (when data doesn't overlap)
- incomplete info (missing data) - 37model sum squares (SSm) (correct)how much variability explained by model
- 38sum of squares residual (SSr) (correct)how much variance cant be explained
- 39mean squares model (Msm) (correct)average variability explained by mode)
- 40mean squares residual (correct)average variability not explained by model
- 41difference between mean squares model (Msm) and model sum of squares (SSm) (correct)Msm - average variance plained by model
SSm - variance not explained for by model
so mean just means average - 42difference between... mean squares residual (Msr) sum of squares residual (SSr) (correct)both variance not explained for by model
but mean squares residual is the average value (as its mean) - 43what does means squares meanaverage variability explained for by experimental manipulation or by other factors
- 44how to interpret coefficient tables that contain B and Unstandardised Bb = ..... [UQ, LQ], UB = ....., t = ...., p = ...
write UB as the weird s looking b - 45which is better for comparing standardised b or unstandardised bSTANDARDised b - as its set in STANDARD deviations, for comparing
Informações
- Descrição
- FlashCards sobre part 2 stats, criado por abi Jane em 09-01-2020.
- Criado por
- abi Jane
- Visualizações
- 2
- Avaliação