How to Analyze Research Data | RISE Research
How to Analyze Research Data | RISE Research
Shana Saiesh
Shana Saiesh
Here is the honest truth about what data analysis is. It is not running a few numbers and calling it a day. It is the part of the research process where you try to determine what your data is, in fact, telling you, as opposed to what you hoped it was going to tell you, and sometimes even as opposed to what it appears to be telling you.
Most students think of it as a formality, while some students who do a good job of it think of it as the heart of the entire project.
Quantitative or Qualitative?
The type of data collected determines all the choices that follow in the analysis.
What is Quantitative Data?
Quantitative data is numerical in nature. It includes survey scores, measurements, test scores, and counts of things. The purpose of quantitative analysis is to identify patterns, compare groups, and determine the statistical significance of the results.
What is Qualitative Data?
Qualitative data is descriptive in nature. It includes interview transcripts, open-ended survey responses, and written accounts. The purpose of qualitative analysis is not to measure the frequency but to understand the results.
Some studies will incorporate both of these. For example, a survey with rating scales and space for text answers is already mixed. An experiment with numerical data and researcher field notes is mixed as well. This is okay and can even be stronger, but it means you are performing two analyses instead of one.
The reason this is important to remember at the beginning is because the tools you will use, the methods you will take, and even the questions you will ask are different based on what kind of mixed study you are performing. Trying to analyze qualitative data in the way you would perform a t-test, or vice versa, will give you results that mean nothing.
Working With Quantitative Data
Start with cleaning. Before any analysis, go through your data and remove incomplete responses, obvious errors, and anything that looks like a mistake. A survey where someone entered their age as 999 is not a data point you want in your analysis. Neither is a duplicate submission or a blank row. This step feels tedious but it changes your results, sometimes significantly.
After cleaning, run descriptive statistics. Mean, median, mode, standard deviation, range. These tell you what your data looks like before you draw any conclusions. Mean works well for normally distributed data. Median is more accurate when your data is skewed or when outliers are present. Standard deviation tells you how spread out your results are around the average. Most student researchers can get through this in Google Sheets or Excel without any specialized software at all.
Then look at relationships. This is where you test whether your variables connect to each other in meaningful ways. A correlation tells you whether two variables move together. A t-test tells you whether two groups differ significantly. A chi-square test works for categorical data, where your responses fall into categories rather than on a numerical scale. You do not need to run all of these. You need to run the one that actually answers your research question.
One thing worth understanding before you move forward: statistical significance and practical significance are not the same thing. A p-value below 0.05 is the standard threshold for significance in most fields, meaning there is less than a 5% chance your result is due to random chance. But a statistically significant result can still be a small or unimportant effect in practice. The number tells you it is real. It does not tell you it matters.
For free tools, JASP is worth knowing about if Excel starts feeling limiting. It is designed specifically for social science statistical analysis, produces publication-ready output, and does not require any programming knowledge.
Working With Qualitative Data
Qualitative analysis is less systematic-looking than quantitative, but it is still a process. And it still requires rigor, just a different kind.
Start by organizing everything in one place. Transcribe any audio interviews. Compile your open-ended survey responses. Gather field notes. You want all your raw data in a form you can actually read and work with before you start.
Then code it. Coding means reading through your data and tagging sections with short descriptive labels. A participant who talks about feeling overlooked in group projects might get tagged "exclusion" or "group dynamics." Someone who describes a teacher who changed how they thought about learning might get tagged "influential educator." You do the same thing across all your data, building a set of codes that describes what is actually in there.
After coding, look for patterns across codes. Which ones appear together? Which ones come up repeatedly? Which ones only appeared in certain types of responses? Those clusters become your themes, and your themes become your findings.
Here is where most students stop short. Describing what participants said is not analysis. Explaining what it means, why those themes emerge, how they connect to your research question and to the existing literature on your topic, that is analysis. The difference between a good qualitative paper and a weak one is almost always this: the weak one summarizes the data, the good one interprets it.
For free tools, Taguette is a simple open-source platform designed for qualitative coding. It works in a browser, requires no installation, and handles the kind of small-scale coding most student research projects involve.
A Few Things That Go Wrong
The correlation-causation problem comes up in almost every quantitative paper written by a first-time researcher. Two variables moving together in your data does not mean one causes the other. They might both be caused by a third variable you did not measure. Or it might be coincidence. A correlation is a starting point for asking why, not an answer to why.
Ignoring outliers instead of explaining them is another common issue. If one participant's response looks dramatically different from everyone else's, deleting it to clean up your results is not acceptable. Investigating why it is different often produces some of the most interesting material in the paper.
In qualitative work, the most common mistake is confirmation bias, finding evidence for what you expected to find and not noticing what contradicts it. The fix is to actively look for negative cases, data that does not fit your emerging themes, and take them seriously rather than treating them as anomalies to be explained away.
And for any type of data: null results are real results. If your analysis shows no significant difference, no meaningful pattern, no clear theme, that is a finding. Writing it up honestly is more valuable than forcing an interpretation the data does not support.
Presenting It in the Paper
Results and discussion are separate sections for a reason. Results report what the analysis showed. Discussion explains what it means. Mixing these two things together is one of the clearest signs of an inexperienced researcher.
For quantitative results, report the statistic, the sample size, and the significance level. For qualitative results, present themes and ground them in direct quotes from participants. Label participants consistently, like P1, P2, or Participant 3, to preserve anonymity while keeping your evidence traceable.
If you are a high school student pushing yourself to stand out in college applications, RISE Research offers a unique opportunity to work one-on-one with mentors from top universities around the world.
Through personalized guidance and independent research projects that can lead to prestigious publications, RISE helps you build a standout academic profile and develop skills that genuinely set you apart. With flexible program dates and global accessibility, ambitious students can apply year-round. To learn more about eligibility, costs, and how to get started, visit RISE Research's official website and take your college preparation to the next level!
FAQs
Q: Do I need to know statistics?
A: For basic quantitative analysis, no. Mean, median, standard deviation, and a basic correlation are enough for most student research papers. JASP walks you through more complex tests without requiring prior statistical training.
Q: What if my results do not support my hypothesis?
A: Write it up honestly. Null results and unexpected findings are legitimate science. Many published papers report them. What matters is that your analysis was sound.
Q: How many participants do I need?
A: For qualitative research, 8 to 12 in-depth interviews can produce rich findings. For quantitative surveys, at least 30 participants are generally needed for basic statistical tests, though more is better.
Q: Can I use AI to help analyze my data?
A: For organizing and summarizing, yes, with caution. For the actual interpretation and write-up, no. The analysis needs to be yours. Using AI to write your discussion section is an academic integrity issue at most journals and institutions.
Author: Written by Shana Saiesh
Shana Saiesh is a sophomore at Ashoka University pursuing a BA (Hons.) in English Literature with minors in International Relations and Psychology. She works with education-focused initiatives and mentorship-driven programs, contributing to operations, research, and editorial work. Alongside her academics, she is involved in student-facing reports that combine research, strategy, and communication.
Here is the honest truth about what data analysis is. It is not running a few numbers and calling it a day. It is the part of the research process where you try to determine what your data is, in fact, telling you, as opposed to what you hoped it was going to tell you, and sometimes even as opposed to what it appears to be telling you.
Most students think of it as a formality, while some students who do a good job of it think of it as the heart of the entire project.
Quantitative or Qualitative?
The type of data collected determines all the choices that follow in the analysis.
What is Quantitative Data?
Quantitative data is numerical in nature. It includes survey scores, measurements, test scores, and counts of things. The purpose of quantitative analysis is to identify patterns, compare groups, and determine the statistical significance of the results.
What is Qualitative Data?
Qualitative data is descriptive in nature. It includes interview transcripts, open-ended survey responses, and written accounts. The purpose of qualitative analysis is not to measure the frequency but to understand the results.
Some studies will incorporate both of these. For example, a survey with rating scales and space for text answers is already mixed. An experiment with numerical data and researcher field notes is mixed as well. This is okay and can even be stronger, but it means you are performing two analyses instead of one.
The reason this is important to remember at the beginning is because the tools you will use, the methods you will take, and even the questions you will ask are different based on what kind of mixed study you are performing. Trying to analyze qualitative data in the way you would perform a t-test, or vice versa, will give you results that mean nothing.
Working With Quantitative Data
Start with cleaning. Before any analysis, go through your data and remove incomplete responses, obvious errors, and anything that looks like a mistake. A survey where someone entered their age as 999 is not a data point you want in your analysis. Neither is a duplicate submission or a blank row. This step feels tedious but it changes your results, sometimes significantly.
After cleaning, run descriptive statistics. Mean, median, mode, standard deviation, range. These tell you what your data looks like before you draw any conclusions. Mean works well for normally distributed data. Median is more accurate when your data is skewed or when outliers are present. Standard deviation tells you how spread out your results are around the average. Most student researchers can get through this in Google Sheets or Excel without any specialized software at all.
Then look at relationships. This is where you test whether your variables connect to each other in meaningful ways. A correlation tells you whether two variables move together. A t-test tells you whether two groups differ significantly. A chi-square test works for categorical data, where your responses fall into categories rather than on a numerical scale. You do not need to run all of these. You need to run the one that actually answers your research question.
One thing worth understanding before you move forward: statistical significance and practical significance are not the same thing. A p-value below 0.05 is the standard threshold for significance in most fields, meaning there is less than a 5% chance your result is due to random chance. But a statistically significant result can still be a small or unimportant effect in practice. The number tells you it is real. It does not tell you it matters.
For free tools, JASP is worth knowing about if Excel starts feeling limiting. It is designed specifically for social science statistical analysis, produces publication-ready output, and does not require any programming knowledge.
Working With Qualitative Data
Qualitative analysis is less systematic-looking than quantitative, but it is still a process. And it still requires rigor, just a different kind.
Start by organizing everything in one place. Transcribe any audio interviews. Compile your open-ended survey responses. Gather field notes. You want all your raw data in a form you can actually read and work with before you start.
Then code it. Coding means reading through your data and tagging sections with short descriptive labels. A participant who talks about feeling overlooked in group projects might get tagged "exclusion" or "group dynamics." Someone who describes a teacher who changed how they thought about learning might get tagged "influential educator." You do the same thing across all your data, building a set of codes that describes what is actually in there.
After coding, look for patterns across codes. Which ones appear together? Which ones come up repeatedly? Which ones only appeared in certain types of responses? Those clusters become your themes, and your themes become your findings.
Here is where most students stop short. Describing what participants said is not analysis. Explaining what it means, why those themes emerge, how they connect to your research question and to the existing literature on your topic, that is analysis. The difference between a good qualitative paper and a weak one is almost always this: the weak one summarizes the data, the good one interprets it.
For free tools, Taguette is a simple open-source platform designed for qualitative coding. It works in a browser, requires no installation, and handles the kind of small-scale coding most student research projects involve.
A Few Things That Go Wrong
The correlation-causation problem comes up in almost every quantitative paper written by a first-time researcher. Two variables moving together in your data does not mean one causes the other. They might both be caused by a third variable you did not measure. Or it might be coincidence. A correlation is a starting point for asking why, not an answer to why.
Ignoring outliers instead of explaining them is another common issue. If one participant's response looks dramatically different from everyone else's, deleting it to clean up your results is not acceptable. Investigating why it is different often produces some of the most interesting material in the paper.
In qualitative work, the most common mistake is confirmation bias, finding evidence for what you expected to find and not noticing what contradicts it. The fix is to actively look for negative cases, data that does not fit your emerging themes, and take them seriously rather than treating them as anomalies to be explained away.
And for any type of data: null results are real results. If your analysis shows no significant difference, no meaningful pattern, no clear theme, that is a finding. Writing it up honestly is more valuable than forcing an interpretation the data does not support.
Presenting It in the Paper
Results and discussion are separate sections for a reason. Results report what the analysis showed. Discussion explains what it means. Mixing these two things together is one of the clearest signs of an inexperienced researcher.
For quantitative results, report the statistic, the sample size, and the significance level. For qualitative results, present themes and ground them in direct quotes from participants. Label participants consistently, like P1, P2, or Participant 3, to preserve anonymity while keeping your evidence traceable.
If you are a high school student pushing yourself to stand out in college applications, RISE Research offers a unique opportunity to work one-on-one with mentors from top universities around the world.
Through personalized guidance and independent research projects that can lead to prestigious publications, RISE helps you build a standout academic profile and develop skills that genuinely set you apart. With flexible program dates and global accessibility, ambitious students can apply year-round. To learn more about eligibility, costs, and how to get started, visit RISE Research's official website and take your college preparation to the next level!
FAQs
Q: Do I need to know statistics?
A: For basic quantitative analysis, no. Mean, median, standard deviation, and a basic correlation are enough for most student research papers. JASP walks you through more complex tests without requiring prior statistical training.
Q: What if my results do not support my hypothesis?
A: Write it up honestly. Null results and unexpected findings are legitimate science. Many published papers report them. What matters is that your analysis was sound.
Q: How many participants do I need?
A: For qualitative research, 8 to 12 in-depth interviews can produce rich findings. For quantitative surveys, at least 30 participants are generally needed for basic statistical tests, though more is better.
Q: Can I use AI to help analyze my data?
A: For organizing and summarizing, yes, with caution. For the actual interpretation and write-up, no. The analysis needs to be yours. Using AI to write your discussion section is an academic integrity issue at most journals and institutions.
Author: Written by Shana Saiesh
Shana Saiesh is a sophomore at Ashoka University pursuing a BA (Hons.) in English Literature with minors in International Relations and Psychology. She works with education-focused initiatives and mentorship-driven programs, contributing to operations, research, and editorial work. Alongside her academics, she is involved in student-facing reports that combine research, strategy, and communication.
Read More
