Research Methods

How hard can it be to draw random samples? Turns out, it's pretty NUTS In #MachineLearning, we're constantly working with probability distributions: from Bayesian posteriors over model parameters to latent spaces in generative models and non-parametric #UncertaintyQuantification distributions We can usually evaluate the likelihood of an individual point, but actually generating representative samples from the landscape is surprisingly difficult, especially when it's multi-modal or high-dimensional I tried out three classic sampling approaches, all sampling the same target probability distribution: 🔵 Random-Walk Metropolis (classic MCMC): Simple to implement, but takes a long time to explore and converge on the distribution 🟠 Hamiltonian MC (HMC): Uses gradients to glide along the probability landscape, covering ground much more efficiently 🔴 NUTS (No-U-Turn Sampler): An extension of HMC that automatically tunes its own trajectory parameters, exploring with minimal manual tuning The smarter samplers reconstruct the distribution quickly, while MCMC takes a while to build up the full picture: smart sampling saves time and gets better answers! #MCMC #DataScience #Intellegens

MICROGLIA ACT DIFFERENTLY IN MALE & FEMALE BRAINS Microglia, the brain’s immune cells, play vital roles in clearing toxins and maintaining neuronal health but can also contribute to neurodegenerative diseases if overactive. New research reveals sex-based differences in how adult male and female microglia respond to the enzyme inhibitor PLX3397, a common tool in microglial research. While male microglia showed the expected depletion, female microglia employed alternative signaling pathways, leading to increased survival. These findings highlight the necessity of sex-specific research in diseases like Alzheimer’s and Parkinson’s, where microglial activity plays a significant role and diagnosis rates differ by gender. This breakthrough emphasizes the importance of tailoring therapies that target microglia based on sex. Further research aims to explore hormonal and inflammatory factors influencing these differences. 3 Key Facts: 1. Sex-Based Microglial Differences:  Male and female microglia respond differently to PLX3397, with females showing increased survival. 2. Neurodegenerative Impact: Findings could reshape how Alzheimer’s and Parkinson’s therapies are developed and studied. 3. Therapeutic Implications: Sex-specific microglial activity may require tailored treatment strategies. Source: https://lnkd.in/gNAQ6mZk

💡 The importance of sex/gender and age disaggregated data in (biomedical & women's health) research When conducting systematic and scoping reviews, I frequently face the challenge of finding data disaggregated by age and/or sex/gender. Many studies either generalize across both sexes/genders or focus narrowly on women of reproductive age (15-49), often neglecting the distinct health needs of teenage girls, premenopausal, or older women. It's time to acknowledge the critical need to incorporate sex/gender and age considerations into our research. Here is why: - Both sex (biological) and gender (sociocultural) factors influence health, disease, and treatment responses. - Age disaggregation is crucial because it enables accurate tracking of health trends across the life course, aids in identifying age-specific risk factors and interventions, and supports more effective health policy planning and program evaluation. - Government/national funding agencies like NIH and the European Commission have begun to implement policies to integrate sex, gender, and, more recently, diversity analysis into the grant proposal process. Integrating sex/gender and age considerations will lead to more rigorous, reproducible, and relevant research - ultimately improving health outcomes for all. Let's discuss how we can drive this important change together. #research #womenshealth #gender

Survey response rates: the most common challenge I hear in demos. But honestly, it’s not the customer’s fault. It’s yours. Let me explain. Example 1: You want to survey people who took a test drive but didn’t buy. You send them an email or SMS. Response rate? Zero. Of course, they’re not your customers yet. Why would they bother? Try this: Don’t send a survey. Call them. Have a real conversation. Example 2: You’re a new bank. Your customers are retired professionals aged 60+. You send them a feedback email. Will they reply? Unlikely. Try this: Use WhatsApp. Then call. (We’ve seen surprising response rates on WhatsApp in this segment.) Example 3: You’re an NBFC, and your customers are in Tier 3/Tier 4 cities. And you send… an email? Try this: WhatsApp. Then call. Example 4: You’re an airline, and you send a survey 2 weeks after the flight. Do you think they even remember the experience? If you want better response rates: --Be in your customer’s shoes --Choose the right channel for your audience --Ask at the right time --Most importantly, don’t let feedback sit in a dashboard. Act on it. And let the customer know. That’s how you earn feedback. Not with reminders, but with respect. #VoiceOfCustomer #ResponseRates #CustomerFeedback #CustomerExperience

Harness the Power of Inferential Statistics! In the last few weeks, I've covered descriptive statistics and correlation matrices - tools that summarize data and reveal relationships. Now, it's time to take the next step: using inferential statistics to make predictions and draw conclusions about entire populations, all from sample data. Inferential statistics allow analysts to: • Test hypotheses and validate ideas with precision • Predict outcomes beyond the available dataset • Save resources while making informed, data-driven decisions An Example in Action Suppose we want to see if implementing a new onboarding program reduces customer churn. By using inferential stats, we can analyze sample results, determine the statistical significance of the impact, and confidently predict its broader effects while accounting for uncertainty. All with the speed and cost-effectiveness of working with a sample of data rather than data collected from our full customer list. Core Principles At its core lie fundamental principles that transform raw data into actionable insights: Random Sampling: This ensures that samples accurately represent the population by minimizing bias. A properly randomized sample gives us confidence in our conclusions and makes generalizations more reliable. Sampling Distributions: This explains how sample statistics (such as the mean) behave across multiple samples. It guarantees that as our sample size increases, our sample data will closely approximate the true population. Confidence Intervals: These offer a range in which we can expect values from our sample to accurately reflect the broader population. A 95% confidence interval indicates that if we were to take multiple samples, the true value would fall within this interval 95 times out of 100. This range allows us to make informed conclusions while acknowledging the presence of uncertainty. Hypothesis Testing: This tests the validity of assumptions about a population. Starting with a null hypothesis, analysts calculate the probability of their result if the assumption is true. This allows us to either accept or reject the hypothesis based on statistical evidence. P-Values: These measure the strength of evidence against the null hypothesis. A lower p-value (typically <0.05) indicates our findings are unlikely to be due to random chance, increasing confidence in our results. These principles aren't just theoretical - they are practical tools for making informed, data-driven decisions. They empower analysts to make generalized conclusions, rigorously test ideas, and confidently predict outcomes. Key Takeaway By understanding inferential statistics, you can work more smarter, not harder. Whether you are conducting hypothesis testing, making predictions, or discovering hidden patterns, these skills are crucial for utilizing data effectively and taking meaningful action.

Your customer survey doesn't just capture feedback on the experience. It IS the experience. Asking at the wrong time can be annoying. For example, many websites use pop-up surveys to ask for feedback. The timing of the pop-up should be carefully chosen to avoid interrupting the customer's workflow. In this example, the survey popped up on the login page. Customers on the login page are likely intent on completing a transaction within their account. The pop-up survey adds unnecessary friction to the process. Even worse? The survey takes a few minutes to complete. Asking customers to stop what they're doing to spend several minutes completing your survey is pretty cheeky. How could you make this experience better? A few ways: 1. Improve survey timing You could offer a survey at the end of a transaction or when a customer's dwell time (time spent looking at one page) exceeded a certain threshold. 2. Simplify your survey Allow customers to share feedback in one click with an optional open comment box. 3. Think twice Do you really need to survey customers here? Only use a survey if you have a clear reason and a specific plan to use the data. LinkedIn Learning subscribers can get more survey tips from my course, Using Customer Surveys to Improve Service. ➡️ https://lnkd.in/eziCufWi Bottom line: Surveys are part of the customer experience. Make sure your survey ask doesn't make the experience worse.

Ever since ChatGPT arrived, there has been a wave of excitement, skepticism, and curiosity about whether - and how - it actually helps students. Now, a systematic review and meta-analysis by Deng et al. in Computers & Education has pulled together findings from 69 experimental studies, shedding new light on what ChatGPT means for teaching and learning. What Did the Research Reveal? 1️⃣ Stronger Academic Performance Studies show that ChatGPT-assisted interventions often lead to higher grades and better written work - especially in language-rich subjects. One caveat? Many experiments did not make it clear whether students were allowed to use ChatGPT during exams, raising questions about genuine mastery versus AI-assisted output. 2️⃣ Positive Motivation - But Mostly for College Learners University students typically felt more engaged and motivated. In K-12 settings, however, the motivational boost was not as pronounced - suggesting a need for age-appropriate strategies and scaffolds. 3️⃣ Perceived Gains in Higher-Order Thinking Learners reported enhanced creativity and critical thinking. The big “but”: most studies relied on self-reports, so future work needs objective assessments (e.g., problem-solving tasks or performance-based measures) to confirm actual skill growth. 4️⃣ Reduced Mental Effort, Uncertain Self-Efficacy ChatGPT may lighten cognitive load - learners felt tasks were less “taxing.” At the same time, studies showed a mixed or non-significant effect on self-efficacy, implying we need a deeper look at whether students gain real confidence or just convenience. What This Means for Educators & Academics? 1️⃣ Design Rich Assessments: To spot genuine skill gains, use project-based tasks that demand application and originality. 2️⃣ Spell Out Tech Policies: Clearly specify whether and how learners can use ChatGPT - especially for graded work. 3️⃣ Look for Long-Haul Impact: Do not just check excitement levels right after introducing ChatGPT; measure whether those positive vibes (and scores) persist weeks or months down the road. 4️⃣ Mind the Methods: If you are studying ChatGPT’s educational impact, conduct power analyses (to ensure you have enough participants) and randomize group assignments to get the most reliable data. This meta-analysis provides early - but promising - evidence that ChatGPT can enrich students’ learning experiences. The next step? Refining the methods, tracking long-term outcomes, and ensuring actual learning gains are assessed - not just AI’s ability to produce polished outputs. Reference: Deng, R., Jiang, M., Yu, X., Lu, Y., & Liu, S. (2024). Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies. Computers & Education, 105224. https://lnkd.in/eXe8agAT

Sex differences in the brain explain less than 1% of gene-expression variation. And yet they might explain why your Alzheimer's drug failed. A new study in *Science* analyzed over one million brain cells from 30 individuals across six cortical regions. The finding that matters most is not the one that makes headlines. Yes, the team identified more than 100 genes with consistent expression differences between male and female brains. But the core quantitative result deserves more attention. Sex accounted for less than 1% of total variation in gene expression. More variation exists within a sex than between sexes. This is not a contradiction. It is the whole point. Small molecular differences do not mean irrelevant molecular differences. They mean context-dependent molecular differences. The kind that modulate disease risk without determining it. The kind that disappear in underpowered studies and reappear in well-designed ones. This matters for anyone working in preclinical neuroscience or drug development. If your study design treats sex as a confounder to control for rather than a biological variable to characterize, you are not reducing noise. You are discarding signal. The field has known for years that schizophrenia, ADHD, and Parkinson's skew male. That Alzheimer's, depression, and anxiety skew female. What was missing was a molecular handle. DeCasien et al. now provide one. Not a definitive mechanism. A starting coordinate. For translational researchers, the implication is operational. Stratifying by sex is not a regulatory checkbox. It is an analytical prerequisite. The 1% that distinguishes male from female gene expression may sit precisely in the pathway your compound targets. Ignoring small effects because they are small is not rigorous. It is the reason 95% of CNS drugs fail in translation. Article in first comment.

5 Effective Ways to Measure Student Progress Tracking student progress goes beyond grades. It’s about understanding how students learn and grow. Here are five key assessment strategies every educator can use: 1. Pre-Assessments Use short quizzes, surveys, or informal discussions before starting a unit to gauge students’ prior knowledge and readiness. 2. Observational Assessments Monitor student behavior and engagement through notes and behavior trackers. These offer real-time insights into their learning journey. 3. Performance Tasks Let students show what they know through projects, presentations, or hands-on activities. These tasks promote creativity and critical thinking. 4. Student Self-Assessments Encourage learners to reflect on their progress using rubrics, checklists, and self-evaluation tools. It builds metacognition and responsibility. 5. Formative Assessments Regular quizzes, exit tickets, writing prompts, and problem-solving tasks help teachers adjust instruction and provide timely support. Why it matters: Using a variety of assessment methods ensures a holistic view of student learning and helps tailor instruction to meet their needs. How do you measure progress in your classroom? #Education #Learning #StudentAssessment #TeachingStrategies #FormativeAssessment #GrowthMindset

Each of these assessment methods brings its own lens to understanding student learning, and they shine especially when used together. Here’s a breakdown that dives a bit deeper into their purpose and power: 🧠 Pre-Assessments • What it is: Tools used before instruction to gauge prior knowledge, skills, or misconceptions. • Educator insight: Helps identify starting points for differentiation and set realistic goals for growth. • Example: A quick math quiz before a new unit reveals which students need foundational skill reinforcement. 👀 Observational Assessments • What it is: Informal monitoring of student behavior, engagement, and collaboration. • Educator insight: Uncovers social-emotional strengths, learning styles, and peer dynamics. • Example: Watching how students approach a group project can highlight leadership, empathy, or avoidance patterns. 🧩 Performance Tasks • What it is: Authentic, real-world challenges that require applying skills and concepts. • Educator insight: Shows depth of understanding, creativity, and the ability to transfer knowledge. • Example: Students design a sustainable garden using math, science, and writing demonstrating interdisciplinary growth. 🌟 Student Self-Assessments • What it is: Opportunities for students to reflect on their own learning, mindset, and effort. • Educator insight: Builds metacognition, ownership, and emotional insight into learning barriers or motivators. • Example: A weekly check-in journal where students rate their effort and note areas they’d like help with. 🔄 Formative Assessments • What it is: Ongoing “check-ins” embedded in instruction to gauge progress and adjust teaching. • Educator insight: Provides real-time data to pivot strategies before misconceptions solidify. • Example: Exit tickets or digital polls that reveal comprehension right after a lesson. These aren’t just data points they’re tools for connection, curiosity, and building bridges between where a student is and where they’re capable of going. #EmpoweredLearningJourney