Stats jokes for no one but myself.
(yeah... outing myself as a fucking nerd...)
ExplainersâŠ.
These are two statistics programs. The better one is jamovi imo, but spss is like fine itâs just like such an ugly interface.
Cronbachâs alpha is a formula used to test the reliability of a measure with multiple parts (like a questionnaire) - and u use this formula to calculate it . A .95 means excellent internal reliability (YUM!). I'm not keen on the alphaverse (i don't get it sorry) but this joke is funny to me :)
Real thing that happened in my supervisors office :(
Yeah i love my data set ok, what about it.
This graph is basically showing statistical power. In order to run an experiment you need a certain amount of power (Power = 1 â ÎČ) to find a certain sized effect. So this is just me saying look itâs the power graph and rhysand is powerful.
Sometimes you have to transform a variable in a data set in order to use it for analysis. This is a screenshot of the transformation page in jamovi, where I just worte a whole load of bullshite about sailor moon in
P < .05 means statistically significant. This means that it is above chance that the effect you found is real, and therefore you can trust your statistic. If it's not than what is the fucking point (sad violin) (it's actually still useful but you know)
Linear relationships between variables are needed for most parametric tests. This joke isnât great but it was an attempt. The joke is that none of these relationships are linear (so i hardly know em!)
Bonferoni corrections can SUCK MY STRAP I hate them. Basically, you know how I just said that a test is only sig if itâs P < .05? Well, when you do multiple tests on the same data set (eg an anova with post hocs), you have to apply a correction to your p value because youâre fishing in the same data pool â which means the likelihood any of the effects you have found are by chance has now been increased by however many tests you have ran. Bonferoni corrections are super conversative (republican statistics lol â I just mean theyâre overly cautious basically, but they are my fav to apply). To do it you divide .05 by the amount of (post hoc) tests you have done on your data set (ie for three post hoc tests on an anova, youâre looking at .05/3 which is .016). Anything above your new sig number is now insignificant, which means 9 times out of 10 your results are now insignificant and the effects you found are just by chance. Itâs heartbreaking every time ok becuase you have to apply this correction and now you have found no effects.
Another Bonfernoi joke
R is a coding language for stats. I hate it. I have to learn it and every time i pull up code i cry and i get tears all over my keyboard and then it's really hard to type because i can't even see the numbers because i'm crying so hard because i hate r so much i hate you r i hate you.
Ok this one is actually really fucking good. You need a normal distribution for your data set for a lot of stats (this is basically like a nice bell curve around the mean - I mean its more complex than that but yeah ! ). When you look at the distribution of your data set you should check if itâs skewed (a lot of the data set is to the left or right of the mean) or if it has high/low Kurtosis (which is the âPeaknessâ or âflatnessâ of the bell curve). If itâs too high or too low in either of these values, youâll have to transform your data set which is FUCKING LONG AF in order to have reliable stats. Therefore â both skew and kurt are two shortenings for stats on distributions. she skew on my kurt until I distribute is like sooooo fucking funny guys I am so proud of myself and itâs a shame because no one will ever care or understand it first time round unless they are chronically online and do stats but itâs like honestly one of the best jokes Iâve ever made.
I really liked #12 so I made it a stamp
