analyzing panel-data in SPSSS using cross-correlation method

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

analyzing panel-data in SPSSS using cross-correlation method

HeleenDW
This post was updated on .
Hi everyone,  

I am having trouble analyzing panel-data in SPSS.

For my master thesis, I want to explore the intermedia agenda-setting effects between Twitter and online news media (the extent to which content transfers from Twitter to news media, and/or vice versa). Therefore I collected tweets and online news articles over the course of one week (mentioning the word 'Trump'). I created a codebook that allowed me to perform a content analysis and measure the frequency of topics (e.g. environment, economics, etc.) reveiving attention during a certain time period, both in online news media (two online newspapers) and in posts on Twitter.

Now I want to analyze the relationship between the media agendas, established in the content analyses. My research question is, whether for example the Twitter posts  at time point 1 (e.g. in the morning) are
correlated with the reporting of online news media at time point 2 (e.g. in the evening).

Many investigating intermedia agenda setting have employed the cross-correlation method, or cross-lagged panel correlation (and Rozelle-Campbell baseline aferwards) in which cross-correlation panels are set up to compare the media agendas at different time points (more specifically: the proportion of attention dedicated to each issue by each news platform during a certain period, in my case six hours).
 
So far, I have entered my data into spss, in which each row is a tweet or news article. I created 4 variables: 1) tweet or online news media outlet, 2) day 3) daypart/time lag (I used a six-hour time lag; four time lags per day: 24:00-06:00;06:00-12:00; 12:00-18:00; 18:00-24:00)  4) category/topic (30 different categories in total). The dataset contains 292 online news articles and 635 tweets.

My problem is that I am not that familiar with SPSS (or statistics in general for that matter), and I don't really know how to precede from now on. If  anyone could give me a hint on how should I analyze my data using cross-correlation method in SPSS, I would be very grateful.

Thank you in advance!

Heleen
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: analyzing panel-data in SPSSS using cross-correlation method

Maguin, Eugene
>> My problem is that I am not that familiar with SPSS (or statistics in general for that matter), and I don't really know how to precede from now on.

Heleen, please find somebody at your college/university and/or in your department to work with you. Depending on what "not familiar" means, you really need somebody you can work with/consult with/talk with one-on-one. Without that level of help, I'm afraid you'll be swallowing water, not swimming, with respect to both spss and statistics. And, you have an adviser. That person has some responsibilities to you and for you. They owe you.

That said. Your dataset looks like this.
Record Article.type date day.period category.1 to category.30
1 tweet 05/24 3 1,0,...
2 news 05/24 3 0,1,..,1,.
Etc for a total of 927 records.

As I understand your question, the analysis is not very hard to do, just correlations. But, your analysis involves lots of data management/manipulation operations to create the datasets for the correlations. Please find somebody to sit with you.

Gene Maguin




-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of HeleenDW
Sent: Wednesday, May 24, 2017 11:51 AM
To: [hidden email]
Subject: analyzing panel-data in SPSSS using cross-correlation method

Hi everyone,  

I am having trouble analyzing panel-data in SPSS.

For my master thesis, I want to explore the intermedia agenda-setting effects between Twitter and online news media (the extent to which content transfers from Twitter to news media, and/or vice versa). Therefore I collected tweets and online news articles over the course of one week (mentioning the word 'Trump'). I created a codebook that allowed me to perform a content analysis and measure the frequency of topics (e.g.
environment, economics, etc.) reveiving attention during a certain time period, both in online news media (two online newspapers) and in posts on Twitter.

Now I want to analyze the relationship between the media agendas, established in the content analyses. My research question is, whether for example the Twitter posts  at time point 1 (e.g. in the morning) are correlated with the reporting of online news media at time point 2 (e.g. in the evening).

Many investigating intermedia agenda setting have employed the *cross-correlation method*, or *cross-lagged panel correlation* (and Rozelle-Campbell baseline aferwards) in which cross-correlation panels are set up to compare the media agendas at different time points (more
specifically: the proportion of attention dedicated to each issue by each news platform during a certain period, in my case six hours).
 
So far, I have entered my data into spss, in which each row is a tweet or news article. I created 4 variables: 1) tweet or online news media outlet,
2) day 3) daypart/time lag (I used a six-hour time lag; four time lags per
day: 24:00-06:00;06:00-12:00; 12:00-18:00; 18:00-24:00)  4) category (30 different categories in total). The dataset contains 292 online news articles and 635 tweets.

My problem is that I am not that familiar with SPSS (or statistics in general for that matter), and I don't really know how to precede from now on. If  anyone could give me a hint on how should I analyze my data using cross-correlation method in SPSS, I would be very grateful.

Thank you in advance!

Heleen



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/analyzing-panel-data-in-SPSSS-using-cross-correlation-method-tp5734247.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: analyzing panel-data in SPSSS using cross-correlation method

Rich Ulrich
In reply to this post by HeleenDW

Here are some problems you face, outside of the SPSS ones for which, as Gene says,

you should find a local advisor.


You have thin data.  If you look at counts for each of your 7x4=> 28 time periods,

you have 20 or so tweets and 10 or so news articles.  That might be okay for one

category of news, but you have 30 categories, so the average is slightly more than

one article per category per period. 


An obvious way to start an analysis is to aggregate, to get counts for each category

for each period.  If you had collected the actual /posting time/, you would have

other options, with the 4-equal-intervals-per-day being only one possible division.

If the counts are cyclic by day, that complicates analyses or their interpretation.

You will probably end up ignoring the 20 or 25 categories that are least used; I would

probably start out by scanning all 30 and dropping most of them as uninformative

because of low counts and low variation.  (If you use "percent of tweets in this

category", you escape some of the cyclicity, in return for giving up the Poisson

assumption for the variability of counts ... if that matters.)


"Content" is potentially problematic.  Was this week typical, with a gaffe or two?

How did you select your data? - This is something to be argued or defended, both

for the choice of tweets and the choice of news items.  If your tweets reflect one

extended argument across days, most data will not have the sort of "peaks" that

could match or mis-match, showing lead-time one way or the other. 


Hope this helps.


--

Rich Ulrich




From: SPSSX(r) Discussion <[hidden email]> on behalf of HeleenDW <[hidden email]>
Sent: Wednesday, May 24, 2017 11:50:40 AM
To: [hidden email]
Subject: analyzing panel-data in SPSSS using cross-correlation method
 
Hi everyone, 

I am having trouble analyzing panel-data in SPSS.

For my master thesis, I want to explore the intermedia agenda-setting
effects between Twitter and online news media (the extent to which content
transfers from Twitter to news media, and/or vice versa). Therefore I
collected tweets and online news articles over the course of one week
(mentioning the word 'Trump'). I created a codebook that allowed me to
perform a content analysis and measure the frequency of topics (e.g.
environment, economics, etc.) reveiving attention during a certain time
period, both in online news media (two online newspapers) and in posts on
Twitter.

Now I want to analyze the relationship between the media agendas,
established in the content analyses. My research question is, whether for
example the Twitter posts  at time point 1 (e.g. in the morning) are
correlated with the reporting of online news media at time point 2 (e.g. in
the evening).

Many investigating intermedia agenda setting have employed the
*cross-correlation method*, or *cross-lagged panel correlation* (and
Rozelle-Campbell baseline aferwards) in which cross-correlation panels are
set up to compare the media agendas at different time points (more
specifically: the proportion of attention dedicated to each issue by each
news platform during a certain period, in my case six hours).
 
So far, I have entered my data into spss, in which each row is a tweet or
news article. I created 4 variables: 1) tweet or online news media outlet,
2) day 3) daypart/time lag (I used a six-hour time lag; four time lags per
day: 24:00-06:00;06:00-12:00; 12:00-18:00; 18:00-24:00)  4) category (30
different categories in total). The dataset contains 292 online news
articles and 635 tweets.

My problem is that I am not that familiar with SPSS (or statistics in
general for that matter), and I don't really know how to precede from now
on. If  anyone could give me a hint on how should I analyze my data using
cross-correlation method in SPSS, I would be very grateful.

Thank you in advance!

Heleen



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/analyzing-panel-data-in-SPSSS-using-cross-correlation-method-tp5734247.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: analyzing panel-data in SPSSS using cross-correlation method

HeleenDW
In reply to this post by Maguin, Eugene
Thanks a lot for your answer and the advice! I am definitely going to sit down with my adviser as soon as possible.

Heleen
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: analyzing panel-data in SPSSS using cross-correlation method

HeleenDW
In reply to this post by Rich Ulrich
Thank you Ulrich for your very helpful answer. Like you suggested, I was planning to eliminate the least frequent categories. I do still have access to the actual posting time of the tweets and articles, so perhaps it will  be useful to consider other options than the four equals intervals per day.
Loading...