Descriptive statistics
DESCRIPTIVES VARIABLES= x1
/STATISTICS=MEAN STDDEV MIN MAX SEMEAN.
Mean values
MEANS tables = x1 by x1a
/Cells = mean count stddev min max.
Split files
sort cases by x1.
split file layered by x1.
descriptives x2
/statistics mean stddev range min max.
split file off.
sort cases country (A).
t-test
t-test groups = y (1 0)
/variables = x.
correlations
correlations y x1.
t-statistic
The t-statistic is calculated by dividing B by Std. Error. Here: 1.068/0.326=3.276.
Missing values
1. Frequencies table to see the missing values
2. In the tables we can see the missing values that are and are not declared as missing
3. 2 ways to declare missing values:
a. Recode
b. Missing values command
c. Multiple values at the same time
, A. BIVARIATE LINEAR REGRESSION
Linear Regression
1. Bivariate regression
regression
/missing listwise
/statistics coeff outs r anova ci
/dependent y
/method = enter x1.
Continuous (Interval/Ratio) IV
The value of the constant term is 21.83. This represents the expected (mean) value of the DV when the
percentage of women in a country’s parliament equals 0. The value for the slope for
“women_parliament”, meanwhile, is 1.07. The coefficient is positive, indicating that the model predicts
that democratic satisfaction is higher in countries with more women in parliament. In particular, we
expect that the percentage of respondents indicating they are very/fairly satisfied with their country’s
democratic performance will rise 1.07 scale points for every 1 unit increase in “women_parliament”.
Bivariate regression with binary IV
The value of the constant term is 57.69. This is the average value of the DV we would expect if/when
“unicameral” = 0. Given the coding of this variable, this is the mean value of “satisfaction” when
“unicameral” = 0 = country has two legislative chambers (bicameral legislature). The coefficient for
“unicameral” is -4.93. This represents a difference in means between bicameral and unicameral
countries. Specifically, we expect that the mean percentage of respondents saying they are
“very”/“fairly” satisfied is approximately 4.93 scale points lower in countries with a unicameral than with
a bicameral legislature.
B. MULTIPLE LINEAR REGRESSION
2. Multiple: just add the other var.
regression
/missing listwise
/statistics coeff outs r anova ci change tol
/dependent y
/method = enter x1 x2
Create interaction term
Compute x1_x2 = x1*x2.
variable labels x1_x2 “x1 x x2”.
execute.
DESCRIPTIVES VARIABLES= x1
/STATISTICS=MEAN STDDEV MIN MAX SEMEAN.
Mean values
MEANS tables = x1 by x1a
/Cells = mean count stddev min max.
Split files
sort cases by x1.
split file layered by x1.
descriptives x2
/statistics mean stddev range min max.
split file off.
sort cases country (A).
t-test
t-test groups = y (1 0)
/variables = x.
correlations
correlations y x1.
t-statistic
The t-statistic is calculated by dividing B by Std. Error. Here: 1.068/0.326=3.276.
Missing values
1. Frequencies table to see the missing values
2. In the tables we can see the missing values that are and are not declared as missing
3. 2 ways to declare missing values:
a. Recode
b. Missing values command
c. Multiple values at the same time
, A. BIVARIATE LINEAR REGRESSION
Linear Regression
1. Bivariate regression
regression
/missing listwise
/statistics coeff outs r anova ci
/dependent y
/method = enter x1.
Continuous (Interval/Ratio) IV
The value of the constant term is 21.83. This represents the expected (mean) value of the DV when the
percentage of women in a country’s parliament equals 0. The value for the slope for
“women_parliament”, meanwhile, is 1.07. The coefficient is positive, indicating that the model predicts
that democratic satisfaction is higher in countries with more women in parliament. In particular, we
expect that the percentage of respondents indicating they are very/fairly satisfied with their country’s
democratic performance will rise 1.07 scale points for every 1 unit increase in “women_parliament”.
Bivariate regression with binary IV
The value of the constant term is 57.69. This is the average value of the DV we would expect if/when
“unicameral” = 0. Given the coding of this variable, this is the mean value of “satisfaction” when
“unicameral” = 0 = country has two legislative chambers (bicameral legislature). The coefficient for
“unicameral” is -4.93. This represents a difference in means between bicameral and unicameral
countries. Specifically, we expect that the mean percentage of respondents saying they are
“very”/“fairly” satisfied is approximately 4.93 scale points lower in countries with a unicameral than with
a bicameral legislature.
B. MULTIPLE LINEAR REGRESSION
2. Multiple: just add the other var.
regression
/missing listwise
/statistics coeff outs r anova ci change tol
/dependent y
/method = enter x1 x2
Create interaction term
Compute x1_x2 = x1*x2.
variable labels x1_x2 “x1 x x2”.
execute.
