Data Analysis Examples Using SPSS OUTPUTS: Significance
Tests
General guideline: the significance value (p or sig.) represent the percentage or the probability that the results are due to chance. If you have p = .10, this mean 10% of your results is due to chance. The convention used in mass communication research is that results must be equal or less than 5% due to chance. That is, p must be smaller than or equals to 0.05 in order to claim the relationship is truly significant.
All of these samples are based on the GSS data set.
Example 1
Research Question: Do respondents of different races have different
beliefs in life after death?
Analysis: The Chi-square statistics is most appropriate in this
case because both the independent variable, race, and the dependent
variable, belief in life after death, are nominal variables. From
the chi-square statistic, it is clear that there is no relationship
between race and the belief in life after death. That is, whites
are not more likely than black or other race to believe in life
after death. (chi-square= 1.079, df = 2, p = .583).
Note: In interpreting the tables, just focus on the chi-square
test table, the row on Pearson Chi-square value, df (degree of
freedom) and Asymp. Sig. (Asymptotic significance). The raw count
table above it helps you to interpret the results. It would be
particularly useful if the chi-square statistic is significant
and you need to explain the relationships of the variables.
Example 2.
Research Question: Do black people more likely to vote for
a black as a president?
Data Analysis: The chi-square statistics is used to examine the
difference between respondents of difference race and their likelihood
to vote for a black president. The significant chi-square value
show that blacks are very likely to vote for black president than
any other races (Chi-square=14.934, df=2, p = .001).
You can recreate a simpler table to demonstrate the difference
in a report:
| Vote for a black president | ||||
| Yes | ||||
| No | ||||
| Total | ||||
2. t-test
Research Question: Do male and female differ in their view
on sex before marriage?
Analysis: Because this research question ask for comparison between
two groups, male and female, and the attitude toward sex before
marriage is an interval data, t-test should be used to analyze
the relationship between sex of the respondents and their attitudes
toward sex before marriage. The GSS data show that male and female
differ significantly in attitudes toward sex before marriage because
the t-value is higher than the critical value of 1.96 at 0.05
significance level. Male respondents are more likely to accept
sex before marriage than female respondents (t = 3.402, df=891,
p = .001). Male respondents have a mean of 2.93 while female respondents
have a lower mean score of 2.43 toward sex before marriage.
Note: In interpreting these tables, focus on the t, df , and significance
value, and ignore other peripheral information such as Levene's
Test of Equality of Variance, standard error difference etc.
You can recreate a simpler table as follows to illustrate your results clearly:
| Male | |||||
| Female |
(1=premarital sex is always wrong, 4=premarital sex is not wrong
at all)
3. ANOVA
Research Question: Do people of difference race have
different attitudes toward sex before marriage?
Data Analysis: One-way ANOVA is used to analyze the difference
in attitudes toward sex before marriage between the three race
groups: 1) white, 2) black, 3) and other races because sex before
marriage is interval data and there are three groups to be compared.
The ANOVA test of the three groups' attitudes toward sex before
marriage show that the three groups do not differ significantly
in their attitudes toward sex before marriage (F (2,890)
=. 766, p = .465). Therefore we can conclude the people of different
race do not have different attitudes toward sex before marriage.
Note: the F(2, 890) is the convention of presenting between group
degree of freedom first and then within group degree of freedom.
The df of between group is 2, and the df of within groups is 890.
In interpreting the tables, you should only focus on the N, Mean,
F ratio value and the Sig. Level, and ignore other peripheral
information such as standard deviation, standard error, sum of
squares between and within groups.
(1=premarital sex is always wrong, 4=premarital sex is not wrong at all)