Summarizing Screening Data_Class 4_post培训资料.ppt

T-test of Inc_Miss, cont’d. unequal variances Use second t-test Any concerns for Inc_Miss? In Practice Many researchers want to use as much of their data as possible but getting a big enough “n” is often a problem therefore methods of preserving data, s.a. Pairwise (or ‘Analysis by analysis’ rather than ‘Listwise’) deletion is often used for missing data in Main Analyses using average values for composite variables, instead of totals Main analyses are conducted with and without “missing data cases”: if results are similar, then the analysis is presented for the full data set (and it is reported that the results are similar for the smaller data set [with non-missing values] ) II. Outliers May be due to Data entry problems “Gaming” Participants in the “tails” of the distribution (who are part of the population) Participants who are not part of the population Outliers can have large effects on your Results – see: “Outliers” at /textbook/statistics-glossary/o/button/o/ Finding Outliers Look at Frequencies for out-of-range data (e.g., for an “8” on a response scale of 1-5) Ensure that Missing value codes have been set up, if req’d. (e.g., “99”) (c) Plot the data: (e.g., examine “Atthouse” using Frequencies) Analyze, Descriptive Statistics, Frequencies, Charts, Histogram, show normal curve “Atthouse” Finding Outliers, cont’d. (c) Test for extreme values using a z-test: Recall that z=1.96 for .05 in the tails To find extreme values, TF recommend a “conservative” critical z of 3.29 (for a total of .001 in the tails): 0 + 3.29 Area = .0005 z - 3.29 To do so … First, convert values of variables into standardized scores then compare with critical Z of +/- 3.29 In SPSS: “Analyze, Descriptive Statistics, Descriptives, Save standardized values as variables”, then “Analyze, Descriptive Statistics, Explore, Statistics, Outliers” the new variable that was created (“Zatthous”) SPSS Output 16 In Practice … You need to decide what to do with any outliers exami