Inferential Statistics Overview

Inferential Statistics T-test, ANOVA and correlation

What is it? What types of questions/hypotheses does it address? How do you do it in SPSS? What does the output look like? How do you interpret it?

T-test • Compares two sample means for significance • Experiments yield two sample means that need to be tested for significance

T-test Study What accounts for different oral proficiency ratings? • ratings on informational interview are dependent variable • student characteristics possible independent variables Theory--applied learning experiences enhance student performance in the classroom Hypothesis—completing an internship will increase ratings on the informational interview

T-test SPSS Steps • • • •

• •

• •

Open SPSS data file Click on ANALYZE Click on COMPARE MEANS Click on INDEPENDENT SAMPLES T TEST Select continuous dependent variable Select nominal grouping variable Define groups (1 and 2) Click OK

Interpreting Output • Check means and group sizes • Look EQUAL VARIANCES ASSUMED • Note t value, df, and significance (two tailed) If significance ≤ .05, report t = value, df = value, p = significance

T-test Output

T-test Narrative Hypothesis—Completing an internship will increase oral proficiency ratings. Result—Completing an internship had no effect on one dimension (invention) of oral proficiency ratings (t = -.028, df = 56, p = .98).

Found no support for hypothesis.

ANOVA • Calculates the differences among means of 2 or more groups • Ratio of variation due to treatment over variation due to error

ANOVA Study What accounts for different written proficiency ratings? • ratings on portfolio are dependent variable • different instructor possible independent variables Theory--instructor ratings should be uniform. Hypothesis—Instructor identity will not influence written proficiency ratings?

ANOVA SPSS Steps • • • • • • • •

• •

Open SPSS data file Click on ANALYZE Click on COMPARE MEANS Click on ONE WAY ANOVA Select continuous dependent variable Select nominal/ordinal factors Click POST HOC TESTS Select TUKEY and significance level Click CONTINUE Click OK

Interpreting Output • Note F ratio, df (between and within), and significance If significance ≤ .05, report F(dfbetween,dfwithin) = value, p = significance Tukey output indicates which groups are different from each other

ANOVA Output

Tukey Output

ANOVA Narrative Hypothesis—Instructor identity will not influence written proficiency ratings. Result—Instructor teaching COM400 had an effect on one dimension (invention) of written proficiency ratings (F (4,100) = 11.98, p = .0001). 1 differed from 3 and 5; 2 differed from 3 and 5; and 3 differed from 4 Found no support for hypothesis.

Correlations • Refers to the extent to which two sets of data are related • Relationships can be direct, indirect or nonexistent

Correlation Study What are you willing to do for a friend? • willingness to offer emotional support to a CSF is the independent variable • willingness to offer instrumental support to a CSF is a dependent variable Theory—Social exchange theory Hypothesis—A relationship exists between emotional support offered a CSF and instrumental support offered a CSF.

Correlations SPSS Steps

Interpreting Output

• • • • •

• Look above or below the diagonal • Note correlation coefficient and significance (two tailed test)

Open SPSS data file Click on ANALYZE Click on CORRELATIONS Click on BIVARIATE Select continuous dependent variable • Select continuous independent variable • Select PEARSON corr coeff, two tailed test and flag significance correlations • Click OK

If significance ≤ .05, report r = value, p = significance and strength of relationship r = ≤ .39, weak relationship r = .40-.69, moderate relationship r = .70-.99, strong relationship

Correlation Output

Correlation Narrative Hypothesis—A relationship exists between emotional support offered a CSF and instrumental support offered a CSF. Result—A relationship exists between emotional support offered a CSF and instrumental support offered a CSF • platonic friend (r = .65, p = .005) moderate • friends with benefits (r = .66, p = .004) moderate Support was found for hypothesis.