0:02 welcome back we will now discuss time
0:05 series as part of your data analysis
0:08 under analyze face time series is a
0:11 sequence of observations over regularly
0:13 spaced intervals of time as mentioned
0:16 the time series chart is used to track
0:18 the performance of your data or your
0:21 process over time example of
0:24 applications are monthly unemployment
0:27 rates for the previous five years daily
0:29 production at a manufacturing plant for
0:32 a month they get by decade population of
0:35 a state of the previous century it can
0:37 also include your SL age or service
0:39 level agreements which are monitored on
0:42 a monthly basis or the number of
0:43 customers that you have served for the
0:46 past weeks on a daily basis or on a
0:49 monthly basis or the number of incidents
0:52 of mishaps in an organization which is
0:56 recorded on a monthly basis as long as
0:58 you are trying to monitor or check the
1:01 performance of your data regardless of
1:05 its type either continuous or attribute
1:08 or discrete you can use time series as
1:10 long as your x-axis is a unit of time
1:13 now let's do on the anatomy of a time
1:16 series chart as you can see here we have
1:18 the y axis which corresponds to a
1:21 continuous variable which is gold price
1:24 on a dollar per ounce and your x axis
1:26 which is here as you can see here there
1:28 are data points that are plotted across
1:31 time and here are the time elements
1:33 remember that you can only use line
1:35 chart when you're using time series not
1:37 using bar child so that's the common
1:39 mistake that visual analysts do commit
1:42 using bar charts when their x axis is a
1:44 unit of time so always use the time
1:45 series chart on those particular
1:48 applications why do we use time series chart
1:49 chart
1:51 generally speaking we use time series
1:52 chart because we want to look for
1:55 patterns or we want to observe how our
1:58 process output behave across time we are
2:00 looking for what we call patterns which
2:02 can be further subdivided into 3 which
2:04 are trend
2:07 shift or cycle let's go with trend a
2:10 trend is a series of data points that is
2:13 increasing or decreasing increasing
2:17 the time decreasing attendance rate
2:19 increasing rejection rate so those are
2:22 possible scenarios that we can see next
2:26 is chef chef means an old performance
2:27 shifting to a new level of performance
2:30 either because of an improvement or
2:34 either because of a problem now as you
2:37 can see here there is a group of data
2:40 points here which shifted downward so
2:42 this is a particular example let's say
2:45 this is the current performance of your
2:47 attendance rate for the past weeks and
2:49 then it suddenly drop or this could be
2:53 the performance of your yield and it
2:55 suddenly drop due to something that
2:58 happened when you detect shift there has
3:00 been a sudden change of an element of
3:02 the process that triggered this
3:06 particular shift next cycle cycle as
3:08 represented by waves like this this
3:10 could also mean a CSUN
3:12 so when you say cycle it's a repetitive
3:15 cyclical movement of your data points
3:17 from one series of time periods to
3:19 another let's say holiday there could be
3:22 holiday season or cycle so those are
3:24 example of cycle though we can have as
3:27 we look on our time series chart now in
3:29 using time series chart this is the
3:31 important patterns that we have to look
3:33 for if in case we found out any of this
3:37 we should ask ourselves what happened to
3:39 our process why is it giving us this
3:42 kind of patterns let's take our first
3:44 example as we use time series flood the
3:46 manager of a shipping yard wants to
3:48 study the amount of cargo that is
3:51 transported the manager collects the
3:53 weight of all cargo that passes through
3:55 the shipping yard each month now we will
3:57 go to our source sheet and copy our data
4:00 shipping to our Minitab so let's look
4:03 for our shipping data and our source
4:06 worksheet just have to copy this and
4:09 then go to our Minitab again you have to
4:11 paste your data on the first row here
4:14 without the CRO numbers now we have two
4:17 columns the month running through more
4:21 than 50 and 100 seven months of data and
4:23 this is the weight of the material ship
4:25 when using time series we can go either stat
4:27 stat
4:30 we have an option here time-series or we
4:33 can go to graph we have an option here
4:35 so for this case we'll go to graph again
4:39 time series plot plot the data in the
4:41 order that it appears in the worksheet
4:45 okay so whatever you are giving the
4:48 Minotaur software it will just plot it
4:51 based on the order so let's go to graph
4:54 and then click time series let's use
4:56 simple because we have one set of data
4:59 here one column and then you just have
5:01 to click your weight data here which
5:03 data columns do you want operate a time
5:05 series plot so double left-click that
5:07 particular column for this case C 2
5:10 double left-click and then you just have
5:11 to click OK
5:14 okay so we have now a time series of
5:18 weight across 107 months so this is your
5:21 y axis corresponding to your rate data
5:24 across 107 months and these are the data
5:27 plotted all over those periods do you
5:31 see any patterns okay so for this case
5:35 we have this lump of data points here
5:38 I think this would mean a shift from
5:44 here to here okay and there's a natural
5:47 cycle as you can see several number of
5:48 trends decreasing trend we have
5:51 increasing trend increasing trend
5:53 decreasing trend those patterns are
5:56 available now what we want to do is to
5:57 understand the behavior of our process
6:01 why it exhibits this patterns and make
6:02 it as a basis for us to further
6:05 investigate now when you're using time
6:08 series for your hypothesis testing or
6:10 root-cause validation you just have to
6:13 state your hypothesis and if your data
6:16 and if you want to prove that using your
6:18 data and check it across time then you
6:20 can use time series let's say is there
6:23 any dependency of your problem - let's
6:26 say the day of the week or week of the
6:28 month now you can use time series to
6:29 prove that so that's the first example
6:31 let's go to the next example the
6:34 administrator of a hospital wants to
6:36 examine the number of courage' patients
6:39 admitted over the past 24 months to
6:41 analyze trends in the
6:44 so for this case we will be using the
6:46 patient data so let's go to our
6:51 worksheet patient and then let's copy
6:55 the data let's go back Minitab create
6:57 new worksheet by pressing ctrl n and
7:03 close this graph now we can paste a data
7:06 so we have two columns again this time
7:09 it's indicated here on the first call of
7:12 the month but this data we only get the
7:16 first 12 months of a year we didn't
7:18 include the second set of twelve months
7:21 and this column corresponds the number
7:23 of cards of patients now what we want to
7:26 do is to create a time series chart
7:33 graph time series simple let's erase
7:35 this and replace with number of patients
7:38 now before you click OK you can actually
7:42 specify the time for scale because you
7:44 have C one month for you to replace the
7:47 index one to twelve we can click
7:50 timescale and then click stamp click on
7:52 this dialog box here and then the
7:55 available stamp column will appear so
7:57 you want to stomp the x-axis with the
7:59 month corresponding to the data points
8:03 and then you can click OK and then click
8:07 ok now we have your time series chart of
8:09 the number of patient cards of patients
8:12 over one year now do you see patterns
8:16 yes there is an increasing pattern here
8:19 from October November and December and
8:22 there is a decreasing pattern here from
8:26 January February March and April there
8:28 could be a possible shift this lump of
8:31 data going to this lump of data and
8:34 another shift here there could be cycle
8:37 if we replicate this one we move to the
8:39 next year we can see cycle now the
8:41 question is why is there this particular
8:44 behavior of this process or output data
8:46 this is number of cards of patients over
8:48 a year you can see that in November and
8:50 December there has been an increase
8:54 until January February and March
8:57 it's because do you have any guess yes
8:59 it's because of the festivities or the
9:03 holiday so this is a data from the US we
9:05 have November December those months
9:08 until March those are festivities we
9:10 just the reason why the number of cards
9:12 of patients should up on this month by
9:14 understanding how the process output
9:17 behave over time can give us insights
9:19 and how we can do our actions or
9:21 improvement actions moving forward
9:23 validating also our claims if let's say
9:27 is months a significant factor on the
9:28 number of courage of patients if that's
9:31 your hypothesis or is the number of
9:33 project patients dependent on the month
9:35 if that's the case then the answer will
9:38 be yes as seen in this particular output
9:40 chart that we have moving forward this
9:42 is how you will validate your hypothesis
9:45 if you have a continuous data or an
9:47 attribute data as your Y and a unit of
Chrome Extension