What Stat Should I Use?
: Description of statistic tests β° 5 minute read
It’s been a bit of a while, hasn’t it? I tried to write several posts in the past few weeks, but I decided to wait on posting them. Anyway, I thought I’d put up some information for my own reference. I hope that you like it too.
| Type of independent variables | Type of dependent variable(s) | Test(s) |
|---|---|---|
| 0 independent variables (1 population) |
||
| interval, normal distribution | one-sample t-test | |
| ordinal or interval | one-sample median | |
| categorical (2 categories) | binomial test | |
| categorical (>2 categories) | Chi-square goodness-of-fit | |
| 1 independent variable with 2 levels (independent groups) | interval, normal distribution | 2 independent sample t-test |
| ordinal or interval | Wilcoxon-Mann Whitney test | |
| categorical | Chi-square test Fisher's exact test |
|
| 1 independent variable with β₯2 levels (independent groups> | interval, normal distribution | one-way ANOVA |
| ordinal or interval | Kruskal Wallis | |
| categorical | Chi-square test | |
| 1 independent variable with 2 levels (dependent/matched groups) | interval, normal distribution | paired t-test |
| ordinal or interval | Wilcoxon signed ranks test | |
| categorical | McNemar | |
| 1 independent variable with β₯2 levels (dependent/matched groups) | interval, normal distribution | one-way repeated measures ANOVA |
| ordinal or interval | Friedman test | |
| categorical (2 categories) | repeated measures logistic regression | |
| categorical (>2 categories) | generalized structural equation modeling | |
| 2 or more independent variables (independent groups) | ||
| interval, normal distribution | factorial ANOVA | |
| ordinal or interval | factorial ordered logistic regression | |
| categorical (2 categories) | factorial (multinomial) logistic regression | |
| 1 interval independent variable | ||
| interval, normal distribution | correlation simple linear regression |
|
| ordinal or interval | non-parametric correlation simple ordered logistic regression |
|
| categorical | simple (multiple) logistic regression | |
| 1 or more interval and/or 1 or more categorical independent variables | ||
| interval, normal distribution | multiple regression analysis of covariance |
|
| categorical | multiple logistic regression discriminant analysis | |
| 2 or more categorical independent variables (independent groups) | 2 or more interval, normally distributed dependent variables | one-way MANOVA |
| 2 or more continuous independent variables | 2 or more interval, normally distributed dependent variables | multivariate multiple linear regression |
| 0 independent variables | 2 or more interval, normally distributed dependent variables | Factor analysis |
| 0 independent variables | 2 or more sets of 2 or more interval, normally distributed dependent variables | canonical correlation |
These are just some of the common tests performed with various combinations of independent and dependent variables. If you don’t know what any of this means, I’ll write up a little glossary soon. It’ll be helpful for me too because sometimes I just go “Oh you know a chi-squared test is what you do in this situation haha” without really thinking about what a t-test actually is.
To be honest most of the time I just do factorial regression or mixed-effects multiple logistic regression… I always have so many variables…
I adapted this table from UCLA’s site, James D. Leeper, Ph.D.’s Choosing the Correct Statistic, and some of my own experience.
Diary stuff
I took the second exam and got 99% on it. Yay!
I got rejected on an apartment application because I just started working my new job and don’t have paystubs yet. Unyay! However, I have started working at the new job so I’m looking forward to doing new and exciting things even if I don’t get paid for it yet. Really, I’m the easy to take advantage of sort where if you tell me that I’ll get to look at interesting data and run analyses on it, I’ll do it for free. I did this for Bobaboard and I did this for AO3 and now I can do it for [REDACTED] now too.
My birthday is coming up and I will be somewhere in my 20s again. Please, if you see a picture of Nanu, send it to me as a gift…
I attended an anime convention and spent a truly horrendous amount of my money on stuff like “A figure of Yoshikage Kira” and “A little plush Kabu doll” and “Lolita fashion that doesn’t feel like it was made out of tablecloth material and has good construction”.
I’d like to get into sewing and garment construction. Hand sewing is a skill I’ll need to polish in the future anyway if my plans to get a PhD pan out and I have to settle for beautiful humble genius surgeon. I used to be really into embroidery when I was a kid but I fell out of practice when I decided to try really hard to be as manly as possible. As you can tell, that did not pan out…
Since this is going up late at night, I think I’ll skip my physics class again tomorrow. I never attend it unless there’s a mandatory quiz… It’s a bad habit but I can use my time better by sleeping1. It’s so early in the morning, and the lectures are posted online the week before they actually happen, so I just watch the video on the weekend. It’s not like the professor answers questions or anything in class, so there’s no difference. I don’t really mind because I learn either way, but I hope that the professor doesn’t hold a grudge against me. They don’t seem to be able to tell Asian students apart according to some of my classmates so maybe they haven’t realized yet…
-
I also already know the material like the back of my hand because I already learned it, but my credit didn’t transfer over, so I’m stuck taking the class for literally no reason… I already learned physics where you use calculus and this is a class where they only use algebra concepts to explain physics like some sort of bad calque. It’s honestly more tedious to try and memorize things based only on algebra rather than using the beautiful math. Why do we have to spend a whole hour explaining how to get what is obviously just the derivative without actually talking about derivatives or limits2… Why do we have to talk about the cross product without matrices or determinants…? It’s just “It’s this way because that’s how it is” sort of boring reasoning. I mean at least we have vectors? But only use 2 dimensional vectors? Even though it’s electromagnetism, which is 3 dimensional? ↩︎
-
Actually, let me scream about this some more. So many fucking times there’s a problem where we’re asked to “find the area under a curve” but we only use linear equations for this so the Approved Formula Method is just the triangle area equation. And I mean… I guess.. But just do the integral! That’s what integration is for! ↩︎
Listening to: Masayoshi Takanaka - δΌθ±ηε€η΄θ±ε£²γ
Today's mood: productive