![](https://crypto4nerd.com/wp-content/uploads/2023/01/1HhvMre5-WJPyoH8M8twFVw.png)
Depression is a major public health concern, affecting millions of people worldwide. One of the key challenges in addressing depression is understanding the impact of different interventions on individuals who suffer from the condition. In recent years, social media platforms have become a powerful tool for studying mental health, as they allow researchers to collect large amounts of data about individuals’ thoughts, feelings, and behaviors.
In this study, we proposed a framework for understanding the impact of social support on individuals who have self-reported a diagnosis of depression on Twitter. We collected data from individuals who used the search tag “diagnosed with depression” to identify users who may be struggling with the condition. We then obtained tweets from these users, as well as tweets from a control group of users who did not self-report a diagnosis of depression.
We used a combination of natural language processing techniques and machine learning algorithms to analyze the data. Specifically, we used term-frequency inverse document frequency (TF-IDF) to calculate the relevance of different terms in the tweets and trained a classifier using a public Twitter dataset to classify tweets as depressive or non-depressive. We then used the BERT model, which takes into account the context of each word in a sentence, to classify the tweets in our main dataset.
To understand the impact of social support on individuals with depression, we calculated a “support level” for each tweet based on the number of likes, replies, and retweets it received. We then compared the support level of tweets posted before and after the intervention (i.e., the point at which the user self-reported a diagnosis of depression) to estimate the treatment effect.
We found that the average treatment effect of the intervention was 1.96, meaning that individuals who received social support on their depressive tweets were less likely to tweet about depression in the future. However, this analysis may be biased by confounding factors such as the individuals’ characteristics and their behavior before the intervention. To address this issue, we used two different statistical matching techniques, Propensity Score Matching (PSM) and Mahalanobis Distance Matching (MDM), to estimate the causal effect of the intervention while controlling for these confounding factors.
PSM is a statistical matching technique that aims to estimate the impact of a treatment by taking into account the covariates that determine whether a person will receive the intervention. We used PSM to match users in the treatment and control groups based on their propensity scores, which were calculated using a logistic regression model with features such as the length of the tweet, the user’s support level, and the user’s characteristics (e.g., number of likes, replies). After PSM, we obtained a new cleaner dataset with less model dependence and more pronounced results, supporting our original hypothesis that social support can reduce the likelihood of individuals tweeting about depression in the future.
MDM is a statistical matching technique that uses standardized data to estimate the difference in means of the treatment effect of both groups, thereby estimating the causal effect. We used MDM to match users in the treatment and control groups based on their Mahalanobis distance, which is a measure of the distance between two points in a multivariate space. We also used a caliper, which is the acceptable maximum distance threshold, to prune any pairs that had a distance greater than the caliper. After MDM, we obtained similar results to those obtained with PSM, supporting our original hypothesis that social support can reduce the likelihood of individuals tweeting about depression in the future.
In conclusion, this study has provided a comprehensive analysis of the impact of social support on individuals who have self-reported a diagnosis of depression on Twitter. By utilizing natural language processing techniques, machine learning algorithms, and statistical methods, we were able to gain a deeper understanding of the relationship between social support and mental health.
Our results suggest that social support can have a positive impact on the mental well-being of individuals with depression. We found that individuals who received support for their depressive tweets were less likely to tweet about depression in the future. This finding is in line with previous research in psychology, which has shown that social support can act as a protective factor against the development of mental health disorders such as depression.
Additionally, our use of propensity score matching and Mahalanobis distance matching allowed us to control for confounding factors and obtain a more accurate estimate of the treatment effect. This is an important step in ensuring the validity of our conclusions.
However, it is important to note that this study has some limitations. Firstly, our sample size was relatively small, with only 300 individuals in the treatment group and 300 individuals in the control group. Secondly, our study relied on self-reported data, which may not be entirely accurate or reliable.
Despite these limitations, this study provides important insights into the relationship between social support and mental health. The use of social media platforms such as Twitter provides an unprecedented opportunity to study mental health at a large scale and in real time. Our findings suggest that social support can have a positive impact on the mental well-being of individuals with depression, and highlight the potential for social media interventions to improve mental health outcomes.
Overall, this study has added to the growing body of literature on the importance of social support in mental health and the potential for social media interventions to improve mental health outcomes. Further research is needed to replicate and expand upon our findings and to better understand the mechanisms underlying the relationship between social support and mental health.
Thanks for reading 🙂
The datasets and the code for this project can be found at — https:// github.com/nmehta32/Causality-between-support-and-depression