Analyzing the Stroop Effect¶
(1) What is the independent variable? What is the dependent variable?
In this dataset, we have taken control of the word/color congruency condition. The independent variable is whether the words shown are congruent with the ink colors or if they are incongruent with the ink colors. The dependent variable is 'Response Time'.
(2) What is an appropriate set of hypotheses for this task? Specify your null and alternative hypotheses, and clearly define any notation used. Justify your choices.
Even though there are a small number of rows in this dataset, we are treating this dataset as the entire population. As a result we are using μ as the population mean.
| Symbol | Explanation |
|---|---|
| $$H_0:$$ | Null Hypothesis |
| $${μ}_{congruent}$$ | The mean of the Congruent population's response times. |
| $${μ}_{incongruent}$$ | The mean of the Incongruent population's response times. |
| $$H_1$$ | Alternate Hypothesis |
$$H_0: {μ}_{congruent} - {μ}_{incongruent} = 0$$
- The Null Hypothesis is: There is no difference between the population means of the two columns (Congruent and Incongruent) for response times. $$H_1: {μ}_{congruent} - {μ}_{incongruent} \neq 0$$
- The Alternative Hypothesis is: There is a difference between the population means of the two columns (Congruent and Incongruent) for response times..
We do not expect the Null Hypothesis to be true. This is the "negative" result. We do expect the mean of the response times for the incongruent columns to be higher. In keeping with standard statistical norms we will make the Null Hypothesis that the means are equal ("negative result"). This will be a two tailed test since the means will be tested diverging from each other in either direction (positive or negative).
The assumptions for these hypothesis are as follows:
- The data is true Stroop data.
- The values in the columns are response time data.
- This is the entire population (not just a sample) that we are dealing with.
- Each row is one individual test that is suitable for pairing (Congruent and Incongruent).
(3) Report some descriptive statistics regarding this dataset. Include at least one measure of central tendency and at least one measure of variability. The name of the data file is 'stroopdata.csv'.
# Import packages
import pandas as pd
import scipy.stats as stats
stats.chisqprob = lambda chisq, df: stats.chi2.sf(chisq, df)
import matplotlib.pyplot as plt
%matplotlib inline
import seaborn as sns
sns.set_style('darkgrid')
df = pd.read_csv('stroopdata.csv')
df.head()
The pandas dataframe .describe method (below) produces the required descriptive statistics. We have the means, standard deviations, and the 5 number summaries (min, max, and quartiles). For central tendency we have the two means (14.0511 and 22.0159). For variability we have the standard deviations (3.5593 and 4.7971). As well for variability we also have the minimum and maximum for both columns.
df.describe()
# Observed difference between means
obs_diff = 22.0159 - 14.0511
obs_diff
(4) Provide one or two visualizations that show the distribution of the sample data. Write one or two sentences noting what you observe about the plot or plots.
# Build the visualizations here
# Histograms for Congruent and Incongruent features
plt.figure(figsize=(8, 5))
plt.xlabel('Response Time')
plt.ylabel('Bin Height')
plt.hist(df['Congruent'], alpha=.5, label='Congruent')
plt.hist(df['Incongruent'], alpha=.5, label='Incongruent')
plt.title('Histogram of Congruent and Incongruent Features')
plt.legend();
From the overlaid histograms you can clearly see that that the entire data set is shifted to the right for the Incogruent Feature. This means that on average the Incongruent Feature's response times are much larger than the Congruent Feature's response times. The histogram's apparent means are also consistent with the df.describe() method that we used above.
df.plot(figsize=(8, 5), title='Paired Test With Response Times on Y Axis', kind='bar');
This bar chart is even more instructive. For every test, it shows that the Incongruent Feature's response time is higher. There are NO exceptions.
df.plot(figsize=(8, 5), title='Paired Test With Response Times on Y Axis', kind='box');
The boxplot is also quite instructive. As you can see Incongruent's data range is consistently higher than all of the similiar values on Congruent (e.g. median, quartiles, min, max). There are two outliers on Incongruent, that are considerably larger than anything else in either Feature.
df.sort_values('Incongruent')
(5) Now, perform the statistical test and report your results. What is your confidence level or Type I error associated with your test? What is your conclusion regarding the hypotheses you set up? Did the results match up with your expectations? Hint: Think about what is being measured on each individual, and what statistic best captures how an individual reacts in each environment.
- We will be using a paired T Test. From http://www.statstutor.ac.uk/resources/uploaded/paired-t-test.pdf This is used to compare two population means where you have two samples in which observations in one sample can be paired with observations in the other sample. This is exactly what we have here with the provided stroop data.
- We will be using a confidence level of .05 in a two tailed test.
- Conclusions will be provided below after we conduct the T Test.
# Perform the statistical test here
t, p = stats.ttest_rel(df.Congruent,df.Incongruent)
print('t-statistic =', t, 'p-value =', p)
- From http://docs.statwing.com/examples-and-definitions/t-test/statistical-significance/ "A t-test’s statistical significance indicates whether or not the difference between two groups’ averages most likely reflects a “real” difference in the population from which the groups were sampled."
- A large t-statistic (and this is large) is typically accompanied by a statistically significant p-value. The p-value here is vanishingly small 4.103e-08 (ergo statistically significant).
- 4.103e-08 <.025 (our level of significance is .05 BUT this is a two tailed test). We would reject the Null Hypothesis and accept the Alternative Hypothesis. There is a signficant difference between the means of the two Features (Congruent and Incongruent).
(6) Optional: What do you think is responsible for the effects observed? Can you think of an alternative or similar task that would result in a similar effect? Some research about the problem will be helpful for thinking about these two questions!
- To make our lives simple, our brains are wired to simply respond to stimuli. We have learned over the years that most stimuli are congruent. When the wind blows the trees move. As a result, it is not necessary to measure the velocity of the wind. You can simply look out the window of the house and if the trees are moving, it is windy.
- When we encounter stimuli that is incongruent, all this pre-set up wiring needs neutralizing for us to properly respond. It takes some time to process the stimuli and respond correctly since we need to fight thru all the pre-conditioning that we have developed over a life time.
- The Stroop effect is all around us. An example would be a police lineup. Supposedly the perp is in the line up. It is congruent that the perp is there. So, the witness picks one of the people in the lineup. Yet, all that may be occurring is the Stroop effect. While this may sound academic, for the person that just got picked out of the lineup for a major crime, it is far from academic.
- The net is for a wide range of stimuli, there is built in congruency bias. This may be helpful for us to run our lives, but ... when true thinking is required, it is a large hurdle for us to a) recognize that is what is occurring and b) continue with the process to think rationally instead of reflexively.