Transformations in Statistics: What are they?

[@trini-rt6xn]

Genius of a video with exceedingly clear and understandable explanations so anyone can understand. 👏 In particular, I enjoyed the "punches" thrown at Type I and Type II Errors! Double BAM! Seriously, well done yet again. ✨

[@pipertripp]

Love your videos and I really like your descriptions. So often, it's hard to see how a particular video fits into the content of the channel, but with this description, it's very easy. Love all the links to relevant content and even to the playlist the programme is in. So helpful!

[@Alias.Nicht.Verfügbar]

bro i just found your channel and I loooove how you combine information with your own sense of humour. Pure gold. Suscribed.

[@LuisRomaUSA]

In economics when we log the x variable we interpret it as a level log model where we interpret x coefficient as 1% increase in x equals coefficient increase in Y

[@frankaragona1463]

Your channel is a gold mine!

[@galenseilis5971]

With zero-inflated data I would consider trying a Tweedie distribution or a mixture of a Dirac delta distribution centered at zero mixed with some other distribution.

[@anyijukamark8257]

i just had a thorough lesson. thank you bro

[@usamasyed2143]

Sir, I think theses playlists are not in sequence. Could please arrange them in sequence? Thanks

[@cedwin4]

So, is it like if we transformation the independent variables, we don't have to do GLM ? Because generally we don't transform the dependent variable and if it is not normal, what to do?

[@abhishekg408]

Just a doubt, isn't the test for normality for residuals. Instead you were visualizing bivariate plots?

[@English_Sauce]

Someone commented that your channel is likea gold mine hahaha. They are right

[@tejasbhagwat877]

Hi Dustin, thanks for these videos. They are great! My question for this one is, does data transformation have the same kind of pros and cons as normalization of data? If not how are they different? Or are these totally different things?

[@QuantPsych]

Same thing (if you're referring to the same "normalization" I know of).

[@galenseilis5971]

The term "normalization" can refer to a variety of things. I will try to answer you under the assumption that you are referring to standardized scores (also termed z-scores), and I will arbitrarily pick preserving monotonicity as one of the pros of transformations. The z-score transformation of a set of data scores is a composition of two operations. The first is the subtraction of the mean of the data scores, which is a monotonic transformation. The second is dividing the result of the previous transform by the standard deviation of the data scores, which is also a monotonic transformation. The composition of two monotonic transformations is itself a monotonic transformation. Thus this form of normalization is a monotonic transformation of the data. As a result, this transformation is translation invariant and positive scaling invariant, so both details of location and scale are lost. en.wikipedia.org/wiki/Standard_score en.wikipedia.org/wiki/Monotonic_function

[@Break_down1]

But if you log transform your outcome, wouldn’t you have to use a log link function instead of identity ?

[@Break_down1]

Eh maybe not. Since it actually just seems to be a trick for getting the data to “look” like it’s normally distributed

[@SalmanAnsari-gj1rq]

I hav a question about interaction plot in logistics regression

[@QuantPsych]

Then you should ask it :) Or subscribe, because I have a video coming out shortly on logistic regression (and I believe I cover interaction plots with logistic).

[@maxwfhde8083]

@@QuantPsych Justin, I also have a question!

[@galenseilis5971]

@@maxwfhde8083 His name is Dustin, not Justin.

[@hypercortical7772]

ty