How do I use the MultiIndex in pandas?

[@infiltr80r]

It's rare to find exactly what you're looking for when it comes to Pandas but this was it.

[@dataschool]

Great to hear! 😄

[@rezas2626]

Slice(None) trick was what I looked everywhere to find and found it here. Thank you so much!

[@dataschool]

You're very welcome! :)

[@annieyang4420]

Life saver video! i have been panic for 2 days , read many pieces and still confused. this helped me out! thank you!

[@dataschool]

You're very welcome!

[@brockobama257]

I couldn't find the Corey Schafer video on MultiIndex so I'm here. And I'm not disappointed... subscribed!

[@dataschool]

Welcome to the channel! 👋

[@ssankarnarayan]

slice(None) instead of : was a very important input. I struggled with it for quite a while before watching this video. Thank you.

[@dataschool]

You're very welcome!

[@owenlie]

24 hours struggling to get the answer, and the answer just single line *unstack()* method. Thank you so much, I'll subscribe immediately!

[@dataschool]

Great to hear!

[@amitsardanayt]

You made this complicated taboo topic so simple to understand. Worth half an hour i spent

[@dataschool]

Thanks for your kind words!

[@cosmopolitanPJ]

Hello from Hungary Kevin! After a period away from the technicalities of python working mostly with Google Analytics and CRO techniques I'm glad to be back to your videos playing around Jupyter. I appreciate a lot the work you do on Data School, so here is a bit of my honest appreciation, thank you and keep it up! John Ostrowski :)

[@dataschool]

Thanks so much John! :)

[@emuratspahic]

Great explanation! Even though I used pandas a lot, I never really understood how multi-indexing worked. Thank you so much for this tutorial, this helped me out a lot!

[@dataschool]

Glad it was helpful!

[@stephenmaddock2863]

Brilliant tutorial. This is the 2nd of yours that I have watched and it covered what I need in my new job. Thank you very much.

[@dataschool]

You're very welcome!

[@raminsadeghnasab9310]

Finallllly, I understand how the multi-index works. Thanks man.

[@dataschool]

You're welcome!

[@khardu]

Thank you for sharing your knowledge and experience beautifully. I am new to Python and Pandas, was struggling to understand how to select a subset of the data in a MultiIndex scenario. You made it a piece of cake by putting it all in one video, and by comparing and contrasting between Series and Data Frame.

[@dataschool]

That's awesome to hear! 🙌

[@billy818]

ive been stuck trying to understand this reading tru the documentation for ages. 10 mins and i get it this vid is great thanks

[@dataschool]

Awesome to hear!

[@shrishailgajbhar3191]

Thanks for helping us out with multiindexing in pandas...Cleared my daunting confusions about it.

[@dataschool]

Happy to help!

[@shadrinan90]

Your lesson is sooo simple and clear. I found everything I wanted to know. Thank you

[@dataschool]

Great to hear!

[@adarshrai22]

This is really a value adding video for data specialists. Thanks a lot for this brother.

[@dataschool]

Glad it was helpful!

[@carolinefawer-wasser2570]

As always great explaination. You are a gifted teacher, so surely well worth the Patreon membership for you.

[@dataschool]

Thank you, Caroline, both for your kind words and for your support through Data School Insiders! :)

[@markcampanelli]

Thanks a lot. Including slice(None) was very helpful.

[@dataschool]

Great to hear - you're very welcome!

[@minglee5164]

Thanks for the excellent tutorial, I have been confused by this issue for a long time.

[@dataschool]

Glad it helped!

[@searchingsoul5910]

Extremely informative, thanks for making it so simple - really good job

[@dataschool]

Thank you! 🙏

[@olenamelnyk2427]

You saved my life with slice(None) (19:45). Thank you so much.

[@dataschool]

You're welcome!

[@mahyarazad]

Thank you gazillion, you are the best! I finally understand the functionality of the Groupby method!

[@dataschool]

Great to hear!

[@donaldbebel7979]

Very good explanation of a confusing topic. Thanks you.

[@dataschool]

Glad it was helpful to you!

[@akashpoudel571]

u r the bestest ever teacher to me.......damn lucid ua guideliness......thanks a tonne sir

[@dataschool]

Thank you! :)

[@florianrudowable]

pandas is awesome and you are too! I cant belive that pandas is free and this video is free. Thank you.

[@dataschool]

Thank you so much! 😄

[@HD_157]

The most helpful video I've ever seen about Python. Thanks a lot!

[@dataschool]

Awesome! Glad it was helpful to you :)

[@DeebzFromThe90s]

Thank you sir. I will shoot you a donation if/when my quant journey presents some alpha. You are a lifesaver.

[@dataschool]

Thank you so much!

[@mohammadkaifi]

This is the solution I was looking for. Thanks!

[@dataschool]

You're welcome!

[@Mo3azSolomon]

Thanks a lot for this simplified Explanation 😍

[@dataschool]

You're very welcome!

[@carlomision9546]

Clear explanation with examples. This is exactly what I needed to know. You have yourself a new subscriber.

[@dataschool]

Awesome! Thanks :)

[@adamyork8407]

This was so cool! I am already thinking of a dozen ways to use multi-indexing in my newbie projects and adding them to my little presentation next week. I owe a lot of my comfort with pandas to you. I hope your holidays went well, Kevin!

[@dataschool]

Awesome, thanks so much for your kind comments, Adam! 👌

[@5a035]

19:50 "slice(None)", I never would have figured that out..... thank you!

[@dataschool]

You're welcome!

[@wiresandwavessolutions4723]

Thanks for making this multi-index topic so easy for all of us. I understood, pivot table perfectly after this video. I am having difficulty in the following : I want to select all Symbols and calculate the difference between the closing price of all SYMBOLs from 4th to 05th .

[@mountainscott5274]

Would you normally go through all of this? It seems like one can just stick to groupby() without complicating the syntax for selection or changing the shape of the dataframe. In other words, have you encountered multi-indexing often in the field?

[@dataschool]

Great question! Personally, I try not to use the MultiIndex very much. That being said, it is still useful to be able to work with it when you have to.

[@wildlife_and_drones1320]

Very helpful. I will be back with questions about machine learning.

[@dataschool]

Great! I've got a free course: courses.dataschool.io/introduction-to-machine-learning-with-scikit-learn

[@shabdarana2848]

Thanks. This topic had been bugging me but you make it seem really simple.

[@dataschool]

It is complex, but glad you are feeling good about it now!

[@TadeuszKurpiel]

time about 19:55 - when it comes to selecting rows according to values of deeper levels of MultiIndex, we can also do it that way: (assumption: 'stocks' has ['Symbol', 'Date'] MultiIndex set up) selection by one value: stocks[stocks.index.get_level_values(1) == '2016-10-04'] and selection by values from list: stocks[stocks.index.get_level_values(1).isin(['2016-10-03', '2016-10-04'])] and when the MultiIndex levels have their names (like here: 'Symbol', 'Date'), we can use these names in the 'get_level_values' method instead of numbers, e.g.: stocks.index.get_level_values('Date")

[@dataschool]

Thanks so much for sharing!

[@TadeuszKurpiel]

@@dataschool ​ - and Reuven Lerner in his film 'Retrieving from a multi-index in Pandas' (time 9min 35 sec) shows one more way - with the 'xs' method: stocks.xs('2016-10-04', level='Date') but I think my way is most flexible one - for example we can make logical expressions like this one (it corresponds to Reuven's dataframe): df[ df.index.get_level_values('Year').isin([1993, 1920, 2010]) & df.index.get_level_values('Sport').isin(['Archery', 'Judo']) & (df.Age < 24.0) ]

[@jsr7599]

Perfect video for what I needed. Thanks!

[@dataschool]

You're welcome!

[@rubencrespo1720]

thank you so much, finally I could change from a Dataframe to a multiindex

[@dataschool]

You're very welcome!

[@coolzkabhijit]

Thanks for the video! I have a question. How to access a particular index without naming the index, like say without naming 'AAPL'

[@ahmadalrashedi5718]

man thank you soooooooooooooo much, I had been stuck on one thing for hours and finally figured it out!! !

[@dataschool]

You're so very welcome!

[@mdfaiz4583]

u make things very simple....easy to understand ....thanks man

[@dataschool]

You're welcome!

[@Brocollipy]

Thanks so much for this video. Really great to see a range of options with such clear advice.

[@dataschool]

You're very welcome!

[@kilocesar]

Great content man! Exactly what I needed!

[@dataschool]

Great to hear! 🙌

[@niva090]

Thank you, this helped me out so much! I have one question however, I hope you can answer for me here: I have a multi index with ID as the first and Day as the second. Now I want to select all ID's that has more than 6 days of data. How can I pass this logical statement to the selector?

[@willd0g]

Excellent - yes a few 💡 moments indeed. Very good

[@dataschool]

Great to hear!

[@umeshdev6403]

Thank you. Nice explanation. how to get max values with the dates

[@ebena.k.8032]

Thank you so much. 'unstack()' has alleviated my headache.

[@dataschool]

You're welcome!

[@zennicliffzennicliff]

Thank you so much. You are such a talented teacher. Where can I find out how I can structurally learn courses from you? Thank you and greetings!

[@dataschool]

Thanks! This post might be helpful to you: www.dataschool.io/launch-your-data-science-career-with-python/

[@mselim-o7p]

Thanks for additional topics. Thanks for your time.

[@dataschool]

You're welcome!

[@reibalachandran4775]

Thanks for the video, I found it very useful and enjoyable to watch :)

[@dataschool]

Great to hear!

[@ramvpage]

Excellent Tutorial ! Thank You!

[@dataschool]

You're welcome!

[@Zielu13]

Thanks. Do you know how to change the background color in dataframe.plot graphs ( the background on which there are: labels and graph name). In subplots you can use 'facecolor' param, but it doesn't work in plots generated from multiindex dataframes.

[@ankitsajwan7520]

Exactly what I was looking for. Thanks :)

[@dataschool]

You're welcome!

[@rayli2221]

Thank you for sharing this information. I would like to ask if I wanted to get this to do portfolio optimization so calculating expected returns and covariance on multi-index dataframe how would you set this up?

[@originalkundukulangara9281]

I am trying to merge from pivot table and although it succeeds, it gives an error. merging between different levels can give an unintended result (1 level on the left,2 on the right) warnings.warn(msg, UserWarning) I tried to check if it is because of multi-indexing, it wasn't. not sure how to resolve it

[@mxmmalinin]

Thank you for sharing your knowledge!

[@dataschool]

My pleasure!

[@CNW21]

As clear as day now , thank you

[@dataschool]

You're welcome!

[@MagicmathmandarinOrg]

Thanks for the thorough explanation. Very helpful.

[@dataschool]

Great to hear!

[@ridthrahman6637]

How to use range selection (:) while inner and outer row labels are inside tuples? Use slice which probably did not allow range selection. Thank you!

[@dataschool]

Sorry, it's hard for me to say off-hand, good luck!

[@pankajthakkar311]

This is awesome, thanks for making this Video Kevin.

[@dataschool]

You are very welcome!

[@twtw785]

hello Kevin, great video, got 2 questions for you. data looks like this "date","time","open","high","low","close" 04/21/82,10:01,528.55,528.55,528.45,528.45 04/21/82,10:03,528.40,528.45,528.40,528.45 04/21/82,10:04,528.40,528.40,528.35,528.40 04/21/82,10:06,528.45,528.45,528.40,528.40 I am trying to do a day HIGH/LOW studies on 1-min S&P futures data. I have 2 questions. 1) How could I print the day HIGH / LOW for each day 2) How could I print the day HIGH / LOW for the first hour of trading I've tried the followings: df.groupby(['date']).high.max() and got this date 1982-04-21 529.80 1982-04-22 530.60 1982-04-23 531.95 1982-04-26 532.80 I am kind of stuck here. I want to have the output displayed all in 2 lines day by day like the following. Date, Time, High Date, Time, Low Could you please help ? thanks a lot in advance. I appreciate your time. data type date datetime64[ns] time datetime64[ns] open float64 high float64 low float64 close float64 dtype: object [e-mail]

[@robertodilisio7535]

Hi Mark! Great video on a very Syntax confusing matter such as multiindex. Thank you very much!

[@dataschool]

You're very welcome!

[@ItsRainingSteak]

Top shelf content right here

[@dataschool]

Thank you!

[@iloveno3]

Thank you. I have learned so much from you already...

[@dataschool]

Great to hear!

[@manikandanpalanivel4941]

I learn something from your video👌but I have some doubts. Kindly explain me for example we are having a issues in our computer in different dates . We need to filter out with O/P as computer type, computer model, windows 10,7,8 , dates in sequence it got fault

[@dangantman5474]

unstack() + reset_index() for the win!

[@venkynew9509]

15:06 indeed… you are eye opener… Also I’m wondering can we able to update cell by condition in side multi index value?

[@包乐-c9f]

Thank you ! Your video is very helpful!

[@dataschool]

You're welcome!

[@amitgaj1]

Thanks Kevin. That's brillant explanation.

[@dataschool]

Awesome! Thanks for your kind words :)

[@nguyenduyta7136]

Hi , thanks for great videos, I have a question , If I have many csv file in one folder each file is data frame of symbol of stocks (APPL.csv, BA.csv, CSCO.csv, MSFT.csv,....) but I just wanna pick a few of them but not all, instead of I use pd.read_csv one by one it's so slow and manually so I would like to use for loop but I still not find out the way. Would you please help me. Thank you

[@SFayed-lg5sp]

it's so helpful!!! thank you so much!

[@dataschool]

You're so welcome!

[@ramutillu5005]

Thanks alot. Very useful. Can u teach how to use multiple groupby in single expression.

[@mounah2182]

Thanks for the video but i have a question how to merge columns and indexes in order to not have blank spaces in headers in my html view ?

[@asdkop7818]

Nice video, anyway I would have liked a final section in which you could have explained the usage of logic over the selection on multi-indexes. Eg: what if I wanted to select data for each symbol and for each date except one? Is it possible to use the same logic as .loc with simple indexes? Thanks :)

[@dataschool]

Thanks for your suggestion!

[@АнтонМакаров-ъ5е]

5:30 - How to combine 2 multiindex columns in one but in different rows? So first row will be AAPL(combined) - Close - Volume. Second row AAPL 2016-10-03 - Close Volume. Third row AAPL 2016-10-04 - Close - Volume and so on???

[@spiderschwein8g]

Saved me hours. Thanks!

[@dataschool]

Great to hear!

[@ocamlmail]

Great explanation, very clear, thank you!

[@dataschool]

Thanks!

[@nonamenoname2618]

Multiindex is very similar to the functionality of the xarray package. And the pandas creators recommend using xarray for multiple dimensions, especially >3. Does anyone has experience with pandas Multiindex vs xarray? What would be the better choice?

[@dataschool]

I haven't used xarray, sorry!

[@thimirabandara679]

Great video! is there a way to must index the column so there is an outer column and inner column? And if so how do you sort that? Because I know sort_column is not a callable function on the datafram

[@fredsarti1]

Why slice(None).... omg. Thank you for your amazing video! Substribed

[@dataschool]

Thanks!

[@abcdefghi2650]

THANK YOU BILLION TIMES!!!!

[@dataschool]

You're very welcome!

[@jerryetc]

Extremely clear, thank you

[@dataschool]

You're welcome!

[@j.p.brochu8592]

Hoping a new version of Pandas will fix the slice(None) necessity and accept the : instead. This is the type of complexity we can spend a lot of time before we find the solution.

[@ckim0717]

Thank you for the videos! I have an additional question on the slice(None) object you made. Let's say you have 100 more symbols and I don't want all of the symbols but want a specific slice of symbols (for example the all symbols after "CSCO" ) how would I do that? stock.loc[("CSCO": , ['2016-10-03', '2016-10-04']), : ] does not work :(

[@willd0g]

Slice(some)

[@marklittlewood2418]

Nice presentation Kevin well done. How about accessing a range of dates for say apple, date:date ?

[@dataschool]

I'm not sure if that can be done using the colon, I'd have to check...

[@Zielu13]

Thanks. Any advice hot to plot multiindex?

[@gggganzo]

Well explined. Would be great a video about data visualization with multiindex data structures also

[@crush3dices]

11:50 dunno what version you are using but for me with pandas 1.1.3 out[21] does not work. I have to use loc[pd.IndexSlice[:,'key'], :] instead or xs('key', level = 1). Though i would prefer the easier way .loc[:, 'key'] but that does not work...

[@UKIP]

Thanks. The aggregation function of the pivot table can be specified as follows: df = stocks.pivot_table(values='Close', index='Symbol', columns='Date', aggfunc=min) df = stocks.pivot_table(values='Close', index='Symbol', columns='Date', aggfunc=max) df = stocks.pivot_table(values='Close', index='Symbol', columns='Date', aggfunc='mean') # the need for the quotes confused me

[@dataschool]

Thanks for sharing!

[@surenderp2863]

very good explanation in this vedio on multi-indexes, could you please explain, why do we need multi levels of columns and multilevels of row labels and again there is droplevel option, could you please explain on this.

[@dataschool]

The short answer is that you should use a MultiIndex if it helps you to represent the structure of the data.

[@jhonatangarcia3826]

Thanks man! You helped me a lot!

[@dataschool]

Great to hear!

[@matuwuhou]

Thanks for the video! Does anyone know if it's possible to select AAPL and 2 dates and MSFT with only one date?

[@marvinjoyceburgstededijk9749]

Great video and many thanks. Greetings from Amsterdam

[@dataschool]

Thanks for watching!

[@szymonk.7237]

Thank you very much for this explanation ❤️

[@dataschool]

You're welcome 😊

[@pankushkukreja3101]

Thanks Alot, really a good series for pandas and easy to learn. I would request if you can make a video on Iteration in pandas like iterrows, itertuples and iteritems. Many Thanks In advance.

[@dataschool]

I think I cover it in this video: kzbin.info/www/bejne/eF7VaomrgJ1jms0

[@АнтонМакаров-ъ5е]

I have a DataFrame very similar to your ser at 5:06. But with many dates in Date column. Every date in Dates has its value. Need to find for every Symbol a date in Dates with the maximum amount of rows. df.groupby( ['Symbol', 'Date']['Value'].count() -works! It seems, all I need is to add a idxmax( )... But no, it gives only one max value, but I need for every group. How to find a date with maximum rows in every Symbol group?

[@kundansinha5603]

How about if the value is not an aggregate and rather an arithmetic 'calculation ' between 2 variables. Ex. Company names for index, date/timestamp for columns BUT for Values it is something like ColumnA/(ColumnB*4.7)?? Thanks :)

[@dataschool]

Not sure, sorry!

[@Zielu13]

Thanks. How to filter multiindex?

[@goncaloaguiar]

I'm having issues with accessing multiindexed dataframe. I do df.set_index with 2 or more columns, followed by sort_index, but when I try to access the dataframe by passing one value of the first level index using .loc function but the following error is returned: 'MultiIndex slicing requires the index to be lexsorted: slicing on levels [0], lexsort depth 0' By the way, my index is a datetime type.