Signal Detection Theory (Intro Psych Tutorial #42)
Vložit
- čas přidán 1. 02. 2017
- www.psychexamreview.com
In this video I explain how signal detection theory relates to psychophysics and the study of absolute and difference thresholds. I also explain how response criteria play a role in signal detection theory and the possibility of type I and type II errors. Finally, I consider applications of signal detection theory in daily life, from detecting dangers to dating.
Don’t forget to subscribe to the channel to see future videos! Have questions or topics you’d like to see covered in a future video? Let me know by commenting or sending me an email!
Need more explanation? Check out my full psychology guide: Master Introductory Psychology: amzn.to/2eTqm5s
Enable closed-captioning or find a full transcript of this video here: www.psychexamreview.com/signal...
One of the finest explanations of SDT. Thank you.
Thanks for commenting!
No one has ever explained this better. Thank you
Thanks!
WHY haven’t I heard of your channel before?! Life saver 🙏🏻😍
Glad you finally found it!
Michael you have saved my sanity this trimester. Blessings be upon you, mate.
Glad to hear that, thanks for commenting!
Great refresher for me, just getting back into neuroscience. Needed to dust up on all of this, and you are a much better teacher than the author of my textbook :D. Thank you!
Glad I can help, let me know if you have any questions!
Thank-you ! You are the Savior ! 😊 Crystal clear explanation ! Please keep making more videos ! It's true pleasure to learn from you.
You're welcome, thanks for commenting, hopefully I'll have time to start making more videos soon!
"Life is noisy💯" Couldn't agree more! Thank you, Sir
Thank you for your video! Its really helpful when studying for the MCAT
You're welcome, best of luck on the MCAT!
Thankyou this explained it so well!!!
The examples are really helpful for conceptualizing this! Thank you!
I'm glad to hear that, thanks for commenting!
You prolly dont give a shit but if you are stoned like me atm then you can stream pretty much all the new series on Instaflixxer. Have been binge watching with my gf for the last few weeks :)
@Walker Colt yea, been watching on InstaFlixxer for since november myself :D
Your explanation was really good & clear. Thanks a lot!
You're welcome!
love your videos they helped me score 95% on my first intro to psych test. thanks :')
You're welcome and congratulations!
Way better than Crash Course videos. You deserve so much more views!!
Thanks!
I just understood Signal Detection Theory, finally. Thank you
You're welcome, glad to hear that!
This was such a good and helpful video, thank you so much!
You're welcome!
Thankyou somuch ...Love your videos it was really really helpful and your examples are really good to understand things in more practical way
I'm glad to hear that, thanks for commenting!
Thank you so much sir for a wonderful discussion......
Absolutely helpful! Thank you so much!
Glad to hear that!
Appreciate your videos so so much for this upcoming Sensation + Perception midterm!
Glad you like them, best of luck on your exam!
@@PsychExamReview 93% woooo! Appreciate that, thank you. Do you use any other social media platforms for your content or just CZcams?
@@Kaaylaac Congratulations, that's great! I have a Twitter (@PsychExamReview) and Facebook page but I'm not very active on those. So CZcams is the main place that I post content.
Love every video!!! Thank you so much for doing this. So helpful. :)
Thanks, I'm really glad to hear that!
You explained perfectly! Thank you very much.
You're welcome!
This helped SO much. Thank you!
You're welcome, I'm glad to hear that!
The way you explain is awesome
Thanks!
Good explanation sir . Keep it up . It was enlightening and informative
Thanks!
you're good at explaining
Thanks!
thankyou for the detailed explanation
You're welcome!
Thankyou sir. This was really helpful. God bless you.
You're welcome, thanks for commenting!
Excellent !!! Thank you sir
You're welcome!
Thank you so much, you are a great teacher!!!
Thanks, glad you liked it!
I was thinking if the type 1 and type 2 errors would differ in individual situations where the null hypothesis is different ?
Thank you so much
Very helpful for me
You're welcome, glad to hear that!
Your analogy in the end is pretty cool. Nevertheless, you're cool in general. Thanks for helping me and the other people.
My pleasure, thanks for commenting!
Helpful ✨ Thank you!
You're welcome!
Thank you very much!
amazing!!! thank you very much
You're welcome!
thanks, understood!
Watching this just 2 hours before my exams is gonna save my grades this sem.
Hope it helps, though I suggest watching a bit earlier :)
@@PsychExamReview Yeah. Actually I did this topic before. Was watching your videos for a good revision.
i guess i wasted money in university. no one could explain it that perfectly! glad to find you here! I've raised a question at the end of this lecture. how can our responses represent to be more effective in making life-risking decisions? i mean how do we know if taking risks is good or not? can signal detection theory help us lead to an optimistic decision?
I'm glad I can help and glad you found my videos too. You ask a great question that I don't think I can fully answer, but signal detection theory suggests that a good starting point when assessing risk is simply to ask whether a miss is better or worse than a false alarm in that situation. Having an idea of this can help to calibrate your response criteria so if you do make an error at least it's the less costly error (though, of course, this can be difficult to assess in reality).
For example, we could imagine someone who has been drinking alcohol who is unsure whether he needs to take a taxi home or is sober enough to drive. Should he take a taxi? In this case a false alarm (could have driven home safely but took a taxi) has a cost in time & money but it's much less costly than a miss (didn't take a taxi when he was too drunk to drive - leading to an accident). For this situation, a good plan would be for the default response criteria to be: if at all unsure, take a taxi. This will mean that he may take a taxi more often than he really needs to, but it greatly reduces the risk of the much more catastrophic error.
@@PsychExamReview aww thank you so much for opening up my mind to this concept. I am literally very thankful to you to make it up very understanding to me through an example. Blessings on your way ☺️
Great explanation, thanks! but how do you "find" a threshold as in psychophysics method? I don't get it yet, for instance, you run a 2AFC task and you want to find a threshold where the stimuli are detected correctly above 70% of the time. It is possible to fit a psychometric curve with SDT?
Great question, I'll attempt to address the concepts, with the caveat that I don't have personal experience doing this kind of research so there are details I'm not knowledgeable of and I don't have expertise in doing these analyses. For an introduction to a 2FC task and psychometric curves, there's a good explanation here: matlaboratory.blogspot.com/2015/04/introduction-to-psychometric-curves-and.html
If we were to translate the "black/white" example from that page to difference threshold, we could use a similar 2FC, presenting two stimuli and asking if they are the same or different (with varying differences between the stimuli on different trials) and create a psychometric curve for accuracy for varying amounts of difference.
For detecting absolute threshold some methods use yes-no (only one stimulus), not 2FC, but others use 2FC for absolute threshold and involve comparing 2 stimuli (such as one time interval with a tone and one without - the participant must decide if the tone was in the first or second interval). Generally, a range of stimuli is presented from no volume to audible volume (one the participant reports detecting nearly every time) and the accuracy detecting these volumes creates a psychometric curve. This wikipedia page has a good summary of several methods for measuring absolute threshold for hearing: en.wikipedia.org/wiki/Absolute_threshold_of_hearing
We could also create additional conditions for presenting the stimuli to see how these influence accuracy and change the curves (similar to the example description of the "black/white" task with different backgrounds).
When we describe reaching 50% accuracy for a threshold, we want to eliminate random chance. In 2FC (or in evenly split Yes-No trials), simply giving the same response every time would lead to 50% correct responses but wouldn't indicate the participant is actually detecting the stimulus or accurately discriminating between the stimuli. To correct for this, we might eliminate 50% of the correct responses as chance guessing. Then we'd want to see 50% accuracy within the remaining trials. In 100 trials for a particular threshold, we could assume 50 of the correct answers are just chance, so we'd want to see 50% accuracy for the remaining 50 trials, meaning another 25 correct responses for a total of 75 correct overall or a "75% correct threshold" (which can also be estimated from the psychometric curve).
I'm not sure if this answered all your questions but I hope it was helpful!
Thanks for your answer, that blog was very useful for understanding 2AFC and the Matlab tutorial was very helpful too. But now my question is more related to if it's possible to combine both methods (Psychophysics and SDT ) to find a threshold, for instance, If I make a 2AFC and I present a constant stimulus (e.g. a light spot) and a variation of the same stimulus in all trials with an increment of 0.1 (i.e. an increment in the size of the light spot), so, I have five different intensities: 0.1,0.2,0.3,0.4 and 0.5 that are presented each of one with the reference stimulus in a random trials and they need to answer where did the biggest spot appears in a screen (e.g. up or down). Then, I find that the threshold is between .2 and .3 but I also notice that my subject is biased to answer "up", that is some kind of bias, right? once a teacher tell us, that you can choose a criterion like, for example, if the participant pressed "up" and in effect the biggest stimulus was there, then you classify that trial as a hit, but if she pressed "up" and the biggest was "down" then you classify as a False Alarm . Now if that is correct, how do I obtain dprime? do I have to group the Hits and FA by the intensities and obtain each dprime individually or do I obtain the general dprime without care what intensities? and finally, how do I obtain the threshold where the stimulus is detected above 75% of the time, taking in account dprime?
Thanks for your help, best regards!
Do you make videos on statistics in psychology or research in psychology ?
I have a playlist on research methods and statistics here. It covers the basics for an intro psych class but not a full methods/statistics course. I'm planning to make a more detailed course on statistics in the future but it will be some time before I can get to that. Hope this helps! czcams.com/play/PLkKvotUGCyLcLVIdlpIYNtHxJ33-JHe5A.html
Thanku sir
very gud video sir
Thank you!
You're welcome!
This was helpful
Glad to hear that!
Amazing
Thanks!
So what are "beta" and "C" in SDT? I understand that d' is the separation between the means of the the signal and noise distributions, but do not understand what beta and C are in this context.
In this context, c and β (Beta) both refer to response bias, or where one’s decision criterion is for when to say yes or no. To the left of the criterion, the response is no, and to the right the response is yes.
c is the distance (in standard deviations) to the criterion measured starting from where the noise and signal distributions overlap. This crossing point doesn’t favor a yes or no response and so this neutral point is c = 0, sometimes called the “ideal observer”. c is negative if the criterion is to the left of this point (more liberal, more false alarms), and positive if it is to the right (more conservative, more misses).
β (Beta) is a ratio of the height of the signal distribution to the height of the noise distribution at the criterion. This provides a likelihood ratio of hits vs. false alarms. Because the heights of both distributions are equal where they cross, at this point β = 1 (neutral; no bias). To the left a false alarm is more likely than a hit (so β < 1 ; liberal), and to the right of this a hit is more likely than a false alarm (so β > 1 ; conservative).
I hope this helps clarify things a bit. These can be confusing concepts so maybe I’ll make a more detailed video on this in the future. Thanks for asking!
Awesome!!!!!!
Mohammad O. Eshkalak Thanks!
wonderful
Thanks!
What is your background? Are you a psychology lecturer ? I'm so intrigued
I earned my bachelor's degree in psychology from Harvard and I teach AP and IB psychology at an international high school in Shanghai.
Great ！！
Thanks!
omg I was thinking of liberal and conservative as the top and bottom boxes and not the left and right and thats why I didn't understand it
WAY better than Khan Academy.
no
I think Type 1 and 2 errors got mixed up!
In this context the Y/N would be for a real effect existing, rather than for the null being true or not, which is another way to present the same concept. This may make it appear flipped but it isn't, you just have to be careful to distinguish whether a chart is referring to the effect or to the null hypothesis (and which is vertical/horizontal, which can also be flipped).
So a Yes response here would be saying there is an effect (rejecting the null) when the null is actually true (there is no effect), leading to a false positive or Type I error, but a No response here would be saying there is no effect (accepting the null) when there is a real effect (null is not true), which would be a false negative or Type II error.
The different ways of presenting this can make it seem more confusing, hope this is clearer!
@@PsychExamReview Yes sir, understood. Thank you so very much!
Are you a professor or a phd scholar?
No, I'm not, I don't have a PhD and I don't work in a university but I've taught psychology at international high schools for about 11 years
@@PsychExamReview Oh that's great! I can clearly see the passion and the experience with which you teach!
@@PsychExamReview I am a cognitive science masters student! Quite an informative channel you have here!
@@hritikgupta4104 Thanks, hope your program is going well!
@@PsychExamReview yes, it is going really well. I'm learning across so many disciplines. I intend to pursue a PhD after this and become a cognitive psychology professor.
this was so helpful, I was wondering if you have a girlfriend?
"I don't have a girlfriend, I just know a girl who would be really mad if she heard me say that." - Mitch Hedberg