This has been answered in part in many places but here I am asking again.
Background: I did calculus and vectors and advanced functions in high school. Then I did a bachelor's degree in social work and a master's in social work and now I'm looking to do a PhD. I consider myself statistically literate, more so than average social work students, and can navigate SPSS. But to conduct rigorous research that is firmly grounded in theory in the social sciences and in the math behind statistics I feel a need for a thorough understanding of mathematics so I can say why I actually did the analyses, why I use certain alpha scores, what I did, and what do the results really mean rather than relying solely on convention and audience ignorance.
What I've gathered so far is I should start with topics in linear algebra, and real analysis… and maybe avoid Discovering Statistics (although it is a favorite among many profs and students)…
but otherwise I am totally lost.
Ultimately, I want to be able to run and understand multi-group confirmatory factor analysis to look at scalar invariance but later in the future also have the flexibility to do some SEM, IRT, Bayesian statistics, and natural language processing.
So for someone looking to self study with no undergraduate math experience….what is my trajectory (e.g. MOOCs, books, get another bachelor's degree)?
BTW I'm going to focus on learning R – for flexibility in the future.
Best Answer
I see various areas you should have a look into:
Here you should understand the most common continuous probability distributions (e.g. normal distribution, t-distribution) and the most common discrete distributions (e.g. binomial distribution and geometric distribution). You should also understand how they are related to each other, e.g. a t-distribution converges to a normal distribution if n goes to infinity. You should also understand concepts like conditional probability and Bayes' theorem and you should have a look into random processes, e.g. random walk.
You should understand the basics of inferential statistics and statistical testing. In statistical testing p-values and power of tests is important.
Linear algebra is one of the most important mathematical concepts for statistics. Important concepts are e.g. the inverse and the transpose of a matrix. You should also be able to calculate with matrices, e.g. multiplication.
There are three different areas of regression analysis: Cross-sectional regressions, panel data and time series analysis. You should go through all of the three areas. Time series analysis might be the most important area of this three areas for practitioners as it is used for forecasting.
After having an overview of the different areas of machine learning you should have look in some of the most common supervised machine learning algorithms (e.g. regression and classification) and the most common unsupervised machine learning algorithms (e.g. clustering, cimensionality reduction and anomaly detection)
R and Python are the most widespread languages for statistical computing. If I were you I would choose R as you need less pre-knowledge in object-oriented computing for using it.