Scribd
Upload
187 views

Tests, Chi-Squares, Phi, Correlations: It's All The Same Stuff

This document discusses the relationships between different statistical tests used to analyze relationships between variables. It explains that t-tests, which compare group means, are equivalent to correlations when the independent variable is binary. Chi-square tests, which analyze relationships between categorical variables, are equivalent to correlations measured by Phi or Cramer's V. Spearman's rank correlation coefficient is also equivalent to Pearson's correlation coefficient when analyzing relationships between ranked variables rather than raw scores. Throughout, it emphasizes that while the statistical tests calculate relationships in different ways, they often lead to the same conclusions about whether variables are significantly related.

Uploaded by

Nejash Abdo Issa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
187 views

Tests, Chi-Squares, Phi, Correlations: It's All The Same Stuff

This document discusses the relationships between different statistical tests used to analyze relationships between variables. It explains that t-tests, which compare group means, are equivalent to correlations when the independent variable is binary. Chi-square tests, which analyze relationships between categorical variables, are equivalent to correlations measured by Phi or Cramer's V. Spearman's rank correlation coefficient is also equivalent to Pearson's correlation coefficient when analyzing relationships between ranked variables rather than raw scores. Throughout, it emphasizes that while the statistical tests calculate relationships in different ways, they often lead to the same conclusions about whether variables are significantly related.

Uploaded by

Nejash Abdo Issa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Newsom

Psy 521/621 Univariate Quantitative Methods, Fall 2020 1

t-Tests, Chi-squares, Phi, Correlations: It’s all the same stuff

In this handout, I provide illustrations of the connection between t-tests and correlations and between chi-
square and correlation.

Correlation equivalents
Correlation is a statistic that describes the association between two variables. The correlation statistic can be
used for continuous variables or binary variables or a combination of continuous and binary variables. In
contrast, t-tests examine whether there are significant differences between two group means. With a t-test, we
have binary independent variable (two groups, which could be coded 0 and 1) and a continuous dependent
variable. If our study is an experiment, then a significant t-test comparing experimental group and control
would suggest that our independent variable has a significant impact (and, therefore association with) the
dependent variable. Significant group differences then imply a correlation between the independent and
dependent variable. The graphs below illustrate that, even if the independent variable has few values, you can
still observe a tendency for the dependent variable to increase in value as the independent variable increases
in value.

t-Tests and Correlations


Below are the results from the t-test handout presented earlier in class, followed by a correlation analysis. The
correlation gives the association between the independent (school type) and dependent variables
(satisfaction).
Independent Samples Test

Levene's Test for


Equality of Variances t-test for Equality of Means
95% Confidence
Interval of the
Mean Std. Error Difference
F Sig. t df Sig. (2-tailed) Difference Difference Lower Upper
satisfaction Equal variances
3.200 .111 -3.000 8 .017 -3.00000 1.00000 -5.30600 -.69400
rating of school assumed
Equal variances
-3.000 5.882 .025 -3.00000 1.00000 -5.45882 -.54118
not assumed

Correlations

satisfaction
Notice that the p-values for the two tests are identical. The
rating of type school significant difference between the means for charter and
school type
satisfaction rating Pearson Correlation 1 .728(*)
public schools is the same as testing whether the school
of school Sig. (2-tailed) .017 type is associated with satisfaction. A special shortcut
N 10 10 formula called the point-biserial correlation used for the
type school type Pearson Correlation .728(*) 1 correlation between a binary and continuous variable is
Sig. (2-tailed) .017
N 10 10
equivalent to the Pearson correlation coefficient.
* Correlation is significant at the 0.05 level (2-tailed).

Chi-square, Phi, and Pearson Correlation


Below are the chi-square results from the 2 × 2 contingency chi-square handout. With SPSS Crosstabs
procedure, you can request Phi (for 2 × 2) or Cramer’s V (for larger than 2 × 2) as a measure of association.
Phi is identical to Pearson’s correlation.
Newsom
Psy 521/621 Univariate Quantitative Methods, Fall 2020 2

Chi-Square Tests

Asymp. Sig. Exact Sig. Exact Sig.


Value df (2-sided) (2-sided) (1-sided)
Pearson Chi-Square 81.940b 1 .000
Continuity Correctiona 80.322 1 .000
Likelihood Ratio 84.606 1 .000
Fisher's Exact Test .000 .000
Linear-by-Linear
81.776 1 .000
Association
N of Valid Cases 500
a. Computed only for a 2x2 table
b. 0 cells (.0%) have expected count less than 5. The minimum expected count is 110.
40.

Symmetric Measures

Value Approx. Sig.


Nominal by Phi .405 .000
Nominal Cramer's V .405 .000
N of Valid Cases 500
a. Not assuming the null hypothesis.
b. Using the asymptotic standard error assuming the null
hypothesis.

Correlations

candidat party voters


candidate party
candidat candidate Pearson Correlation 1 .405**
Sig. (2-tailed) .000
N 500 500
party voters party Pearson Correlation .405** 1
Sig. (2-tailed) .000
N 500 500
**. Correlation is significant at the 0.01 level (2-tailed).

The significance tests for chi-square and correlation will not be exactly the same but will very often give the
same statistical conclusion. Chi-square tests are based on the normal distribution (remember that z2 = χ2), but
the significance test for correlation uses the t-distribution. With large sample sizes (e.g., N > 120) the t and the
normal z-distributions will be the same (or, at least, extremely close). 1

Correlation and Spearman's Rank-Order Correlation (Rho)


A third type of correlation equivalent is one that is named for a correlation between two sets of ranks. For
example, if I had 20 products lined up and I asked you to rank them from 1 to 20. I could then examine how
these ranks were related to another set of ranks, say for a second type of product. The association between
the two sets of rank scores is described by the Spearman's rho. Notice the correlation coefficient and the p-
value are exactly the same for the two analyses (Pearson’s correlation on the left and Spearman’s rho on the
right).

1
This "approximate significance" is reported by most researchers regardless of the sample size. Perhaps this is because, for smaller sample sizes,
there are not any simple recommendations. There does not seem to be much consensus in favor of the Yates continuity correction, and the Fisher
Exact test, which is more powerful, depends on assumption that the marginal frequencies are fixed (determined values as in an experiment).

You might also like