Skip to main content

How do I carry out regression analysis with a sample size of only 28 and number of variables (including DV) 14?

For regression, this is way too less number of samples. It is advisable to use 20–30 samples par variable. Therefore, what you can do here, is consider the correlation of the independent variables with the dependent variables and choose the highest correlated variable to the dependent variable and build a one variable model once. You can also try out principal component analysis on the sample to create 3 effective variable to capture the variance mostly, but I suppose the effectiveness of the PCA is also not that good in a sample size of yours. I wonder whether your main task is to do regression or not. Because such small data are seen mostly in neuroscience and psychology where the main task is to find out underlying factors and not doing any prediction of sort. If you are also having similar reasons, then resort to tests like ANOVA, mANOVA, rank tests and others and devise them carefully enough to find out the effects you are trying to find out.

Finally, if you have a cost effective way to increase your data sample (remember, the data should ideally be independent of each other) then please do increase your data at least upto 500 or so. And then only you can see some good results.

For small samples, I found these links which you may find interesting to read:

(1) When small samples are problematic - Eiko Fried

(2) Problems with small sample sizes

Also, if you are interested in linear regression in general, please follow linear regression  in my blog.

Comments

Popular posts from this blog

Tinder bio generation with OpenAI GPT-3 API

Introduction: Recently I got access to OpenAI API beta. After a few simple experiments, I set on creating a simple test project. In this project, I will try to create good tinder bio for a specific person.  The abc of openai API playground: In the OpenAI API playground, you get a prompt, and then you can write instructions or specific text to trigger a response from the gpt-3 models. There are also a number of preset templates which loads a specific kind of prompt and let's you generate pre-prepared results. What are the models available? There are 4 models which are stable. These are: (1) curie (2) babbage (3) ada (4) da-vinci da-vinci is the strongest of them all and can perform all downstream tasks which other models can do. There are 2 other new models which openai introduced this year (2021) named da-vinci-instruct-beta and curie-instruct-beta. These instruction models are specifically built for taking in instructions. As OpenAI blog explains and also you will see in our

Can we write codes automatically with GPT-3?

 Introduction: OpenAI created and released the first versions of GPT-3 back in 2021 beginning. We wrote a few text generation articles that time and tested how to create tinder bio using GPT-3 . If you are interested to know more on what is GPT-3 or what is openai, how the server look, then read the tinder bio article. In this article, we will explore Code generation with OpenAI models.  It has been noted already in multiple blogs and exploration work, that GPT-3 can even solve leetcode problems. We will try to explore how good the OpenAI model can "code" and whether prompt tuning will improve or change those performances. Basic coding: We will try to see a few data structure coding performance by GPT-3. (a) Merge sort with python:  First with 200 words limit, it couldn't complete the Write sample code for merge sort in python.   def merge(arr, l, m, r):     n1 = m - l + 1     n2 = r- m       # create temp arrays     L = [0] * (n1)     R = [0] * (n

What is Bort?

 Introduction: Bort, is the new and more optimized version of BERT; which came out this october from amazon science. I came to know about it today while parsing amazon science's news on facebook about bort. So Bort is the newest addition to the long list of great LM models with extra-ordinary achievements.  Why is Bort important? Bort, is a model of 5.5% effective and 16% total size of the original BERT model; and is 20x faster than BERT, while being able to surpass the BERT model in 20 out of 23 tasks; to quote the abstract of the paper,  ' it obtains performance improvements of between 0 . 3% and 31%, absolute, with respect to BERT-large, on multiple public natural language understanding (NLU) benchmarks. ' So what made this achievement possible? The main idea behind creation of Bort is to go beyond the shallow depth of weight pruning, connection deletion or merely factoring the NN into different matrix factorizations and thus distilling it. While methods like knowle