Did you read the paper? I don't see how you could walk away with the impression that this is "just correlation", let alone bullshit. Moreover, your quip about n=60 strikes me as the kind of comment made by people who don't know how research and statistics work.
n=60 is actually pretty big for this kind of study (most cognitive studies involving fMRI, for instance, hover around n=20). Either way, simply looking at the sample size is deceiving, as statistical power is what matters in the end. If sufficient power can be achieved with 60 persons, then yours is not a valid criticism.
Concerning the design, this is a straight-up experimental manipulation and not just a correlative study. There were two groups, one which received real tDCS and the other which received a sham tDCS. Subjects were assigned to each group in a double-blind fashion, and the measure of interest here is the improvement between the first test (pre manipulation) and the second test (post manipulation). The use of a sham rules out any placebo/nocebo effects.
This is exactly the kind of study that one would design in order to test causality. We have both temporal precedence and covariance; together, these strongly suggest that tDCS is responsible for the observed effects.
A correlation in your sample (especially at that size) does not imply a correlation in the population, even if you validate it with additional sham tests.
The study performed statistical hypothesis testing to determine that the means of the two groups are probably different.
Edit: and more to your point, correlation of the sample implies correlation of the population if the sample is representative. Is this the case for this particular paper? I don't know and neither does anyone else; that's why replication is the gold standard of scientific validity.
> A correlation in your sample (especially at that size) does not imply a correlation in the population
At least quote the article correctly.
> That is, correlation in the data you happen to have (even if it happens to be “statistically significant”) does not necessarily imply correlation in the population of interest.
(Emphasis mine)
Otherwise you might as well throw out centuries of mathematical, statistical and scientific progress. That laptop of yours? Throw it in the trash as it is the culmination of thousands of sham tests and correlations and is therefore clearly impossible.
It's getting really annoying how many self-annoited gods of statistics keep creeping up online.
60 is plenty if the effect is large enough. Hey, even n=4 can make for a good study. Give two of them cyanide and two of them a placebo, let all the trial-size-too-small-correlation-isnt-causation-people watch and then ask them to back up their criticism with a nice helping of the tested substance.
Perhaps the sample size isn't as large as you'd prefer, but the use of causal language is appropriate here. Indeed, the authors applied a causal intervention (tDCS or a sham treatment) to randomly assigned subjects. This type of study design is pretty much the classical way to infer a causal relationship.
Who said you need more persons for a study like this?
If the sample is representative (and they don't even claim ot attempt to study something that would only affect a particular group) then 60 are fine, in fact they could do with even less.
Besides, what correlation? This is direct observation, there's no other parameter in play in the test setup.
n=60 is actually pretty big for this kind of study (most cognitive studies involving fMRI, for instance, hover around n=20). Either way, simply looking at the sample size is deceiving, as statistical power is what matters in the end. If sufficient power can be achieved with 60 persons, then yours is not a valid criticism.
Concerning the design, this is a straight-up experimental manipulation and not just a correlative study. There were two groups, one which received real tDCS and the other which received a sham tDCS. Subjects were assigned to each group in a double-blind fashion, and the measure of interest here is the improvement between the first test (pre manipulation) and the second test (post manipulation). The use of a sham rules out any placebo/nocebo effects.
This is exactly the kind of study that one would design in order to test causality. We have both temporal precedence and covariance; together, these strongly suggest that tDCS is responsible for the observed effects.